This article was adapted from one of our previous virtual FP&A Summits, featuring Amit Kurhekar, Head of Data at MoneyLion.

Unless you’ve been consistently offline over the last few years, you’ll know that the financial industry is undergoing a significant transformation driven by AI and machine learning technologies.

This revolution isn’t just about adopting new technologies but about changing how financial services and processes are delivered and experienced by consumers.

In this article, we’ll explore some of the most compelling AI and ML strategies in finance with use cases to show how they work in real-life scenarios.

Whether you’re a financial professional or simply interested in the evolving landscape of FinTech, this article offers valuable insights into the intersection of finance, AI, and digital transformation.

Case study: Day in the life of ‘financially savvy’ John

Let me introduce you to John. He considers himself to be very financially savvy, he’s in his 30s, intelligent and he uses a smartphone like so many of us.

One day, he receives a notification on his phone that reads:

“John, your utility bill of $50 is due tomorrow. Do you want to pay now?”

A few seconds later, another notification comes through,

“John, your net-worth increased by 1% last week with Apple stock making the maximum gains.”

John gets on with his day. He goes to work, enjoys chatting to his co-workers, and then in the afternoon, he notices yet another notification on his phone. This one says,

“John, you have excess balance in your savings account. Invest 20% of the amount to earn an extra 8% vs keeping in your savings account. Invest now?”

These are smart notifications and nudges and in today’s financial world, it’s a reality. If you’re not using technology to help improve your finances, you’re missing out. 

By embracing AI and ML, you can make a huge impact not just in your role but also in your daily life.

Pillars of digital transformation 

Within digital transformation, there are emerging technologies. Most companies are utilizing these emerging technologies to drive and improve consumer experiences. These include things like internet of things (IoT), robotics, AR/VR and Cloud.   

Before 2020, not many people were working online or working from home, and then almost the majority of the IT workforce moved into remote working. The transformation from almost everyone working in-office to everyone working remotely because of Covid meant that many people had to embrace technology in new ways. There was a huge mobilization of IT and IT infrastructure. 

I think that both AI and ML are critical pieces that are enabling today’s world. So, a part of that could be coming as simple as receiving smart nudges throughout the day on your smartphone or you could even have nudges to help you forecast numbers for your financial forecast.

Cloud computing is the main support for many companies worldwide. More businesses are moving to cloud computing to improve how they work and compete.

It’s important to find the top security problems in cloud computing. Data leaks caused by cloud setup mistakes and past data leaks need to be watched. This is to avoid impact on the company. 

What is cloud computing? 

Cloud computing changes how we manage, access, and store data. This is done through internet services. This is different from the old way of using storage devices.  

The new cloud-computing model means you do not have to worry about managing servers. Both companies and people can benefit. They get strong data security and flexible, low-cost, and easy-to-adapt data solutions in the cloud. 

Why do you need cloud computing? 

Companies can use secure data centers, lower infrastructure costs, and do operation optimization at full length. It increases efficiency, lowers costs, and empowers businesses. 

With cloud computing, an organization can quickly adjust resources to match demand without requiring large initial hardware investments. An organization needs to pay for only the resources it consumes, lowering expenses for infrastructure and upkeep. You can access data and applications remotely with an internet connection, which increases accessibility to work and collaboration. You can, however, enable fast deployment of new applications and services, eliminating the lengthy lead times of traditional IT methods. In cloud computing, service providers take care of maintenance and updates, guaranteeing that you constantly receive the most up-to-date features and security. Numerous cloud services provide strong backup and recovery options, reducing downtime in the event of data loss. It streamlines IT resource management, enabling teams to concentrate on strategic projects instead of daily upkeep. 

Cloud security issues

There are multiple security issues in cloud computing, and there are hurdles to overcome to secure data and still be able to maintain operational reliability. In this article, we explore the main security concerns in cloud computing and the extent to which they could possibly harm businesses. 

Data loss

Data leakage has been a serious issue in cloud computing. Well, that is true, but only if our sensitive data is being taken care of by someone else whom we cannot trust absolutely, and just the opposite.

Therefore, if cloud service security is going to be baked by a hacker, then hackers can surely get a hold of our sensitive data or personal files. 

Insecure APIs

APIs are the easiest way to talk about the Cloud and need protection. Moreover, as third-party access public clouds, they too can be vulnerabilities to a cloud.

To secure these APIs, implementing SSL certificates is crucial, as they encrypt data in transit, making it harder for hackers to intercept sensitive information. Without this layer of security, attackers can exploit weaknesses in the API, leading to unauthorized access or data loss.  

Account hijacking

The most serious and pressing security threat out of myriads of cloud computing is account hijacking. Once a hacker compromises or hijacks the Account of a User or an Organization, he can access all unauthorized accounts and other activities. 

Change of service provider

Change of service provider is also an important Security issue in Cloud Computing. Many organizations will face different problems like data shifting and different charges for each vendor while shifting from one vendor to another. 

Skill gap 

The biggest problem with IT Companies that do not have skilled Employees is the need to shift to another service provider while working, another feature required, how to use a feature, and so on. Therefore, it requires an extremely skilled person to work in cloud computing. 

Insider threat

On the face of it, this would come out unlikely, but in reality, cloud security threats are those insiders that pose a serious threat to the organizations that avail cloud-based services.

These persons with authorized access to the most needed company resources may indulge in some forms of misconduct, either intentional or unintentional, which will lead to the misuse of their sensitive data. Sensitive data will include client accounts and all critical financial information. 

The important fact to be considered is that the threats from within in cloud security are likely to come through either malicious intent or unintended and just plain negligence. Most such threats can mature into serious violations of security if they develop further and can thereby put sensitive data at risk.

To fight effectively such insider threats while maintaining, at the same time, the confidentiality of data being protected and stored in the cloud, access control must be proper, along with tight and strict access controls.  

Moreover, full training courses including minute details about security should be provided to every member of the staff. In this regard also, monitoring should be done periodically. It is these aspects that have been the main reasons for protection against internal threats that may go about happening. 

Malware injection

The most potent cloud security threats are malware injections. Evil code is concealed in the guise of legitimate code in cloud services. The attacks compromise data integrity because malignant options allow attackers to eavesdrop, modify information, and escape data without detection.

It has become essential to secure the data from eavesdropping in cloud computing and security is an essential aspect. This has become a serious threat to the security of the cloud environment; it should be counter-attacked through careful vigilance and robust security to avoid access to the cloud infrastructure.

Misconfiguration

Indeed, misconfigurations in cloud security settings have proved to be one of the leading and most common causes of data breaches in the present-day digital, and these incidents are mostly the offspring of less-than-perfect practices about managing an effective posture of security.

The user-friendly nature of cloud infrastructure, set up primarily to allow easy exchange and interaction of data, poses significant hurdles to directing access of the data to only a targeted entity or personnel. 

Data storage issue 

This distributed cloud infrastructure is spread all over the globe. Sometimes it tends to keep user data outside the jurisdictions of the legal frameworks of certain regions, raising the range of such data among local law enforcement and regulations. The user dreads its violation because the notion of a cloud makes it difficult to identify one server in the process of transferring data overseas. 

Shared infrastructure security concerns

Multi-tenancy is the sharing of resources, storage, applications, and services from one platform with many at the cloud provider’s site. This tends to enable the provider to recoup high returns on investment but puts the customer at risk. Hence, an attacker can use multi-homing options to make a successful attack against the remaining co-tenants. This has a privacy problem. 

Conclusion 

The business world is changing rapidly, and the rise of cloud computing has created huge security and privacy concerns. In the cloud, there are many issues, such as multiple users sharing the same infrastructure and relying on third parties. These make data vulnerable.

Organizations must be proactive to protect data. They need strong encryption, controlled access, regular security audits, and a clear understanding of their shared responsibility with cloud providers. 

Data pipelines are critical for organizations handling vast amounts of data, yet many practitioners report challenges with responsiveness, especially in data analysis and storage.

Our latest generative AI report revealed that various elements within the pipeline significantly affect performance and usability. We wanted to investigate what could be affecting the responsiveness of the practitioners who reported issues. 

The main area of data workflow or pipeline where practitioners find the least responsiveness is data analysis (28.6%), followed by data storage (14.3%) and other reasons (14.3%), such as API calls, which generally take a significant amount of time.

What factors have an impact on that portion of the data pipeline?

We also asked practitioners about the factors impacting that portion of the pipeline. The majority (58.3%) cited the efficiency of the pipeline tool as the key factor. This could point to a pressing need for improvements in the performance and speed of these tools, which are essential for maintaining productivity and ensuring fast processing times in environments where quick decision-making is key.

With 25% of practitioners pointing to storage as a significant bottleneck after the efficiency of the pipeline tool, inadequate or inefficient storage solutions can impact the ability to process and manage large volumes of data effectively. 

16.7% of practitioners highlighted that code quality disrupts the smooth operation of AI pipelines. This can lead to errors, increased downtime, and complicated maintenance and updates. 

Code quality

The quality of the code in the data pipeline is key to its overall performance and reliability. High-quality code often leads to fewer errors and disruptions, translating to smoother data flows and more reliable outputs. 

Examples of how high code quality can enhance responsiveness:

1. Error handling and recovery2. Optimized algorithms 3. Scalability4. Maintainability and extensibility5. Parallel processing and multithreading6. Effective resource management 7. Testing and quality assurance

Efficiency of pipeline tool

Efficient tools can quickly handle large volumes of data, helping to support complex data operations without performance issues. This is an essential factor when dealing with big data or real-time processing needs, where delays can lead to outdated or irrelevant insights. 

Examples of how the efficiency of pipeline tools can enhance responsiveness:

Data processing speed Resource utilizationMinimized latencyCaching and state managementLoad balancingAutomation and orchestrationAdaptability to data volume and variety

Storage

Storage solutions in a data pipeline impact the cost-effectiveness and performance of data handling. Effective storage solutions must offer enough space to store data while being accessible and secure. 

Examples of how storage can enhance responsiveness:

Data retrieval speedData redundancy and backupScalabilityData integrity and securityCost efficiencyAutomation and management toolsIntegration capabilities

What use cases are driving your data pipeline?

What use cases are driving your data pipeline?

We also asked respondents to identify the specific scenarios or business needs that drive their data pipelines’ design, implementation, and operation to understand the primary purposes for which the data pipeline is being utilized within their organizations.

Natural language processing, or NLP, was highlighted as the main use case (42.8%), with an even distribution across the other use cases. This could be due to businesses increasing their operations in digital spaces, which generate vast amounts of textual data from sources like emails, social media, customer service chats, and more.

NLP

NLP applications require processing and analyzing text data to complete tasks like sentiment analysis, language translation, and chatbot interactions. Effective data pipelines for NLP need to manage diverse data sources like social media posts, customer feedback, and technical documents.

Examples of how NLP drives data pipelines:

Extracting key information from text dataCategorizing and tagging content automaticallyAnalyzing sentiment in customer feedbackEnhancing search and discovery through semantic analysisAutomating data entry from unstructured sourcesGenerating summaries from large text datasetsEnabling advanced question-answering systems

Image recognition

Image recognition analyzes visual data to identify objects, faces, scenes, and activities. Data pipelines for image recognition have to handle large volumes of image data efficiently, which requires significant storage and powerful processing capabilities. 

Examples of how image recognition drives data pipelines:

Automating quality control in manufacturingCategorizing and tagging digital images for easier retrievalEnhancing security systems with facial recognitionEnabling autonomous vehicle navigationAnalyzing medical images for diagnostic purposesMonitoring retail spaces for inventory controlProcessing satellite imagery for environmental monitoring

Image/visual generation

Data pipelines are designed to support the generation process when generative models are used to create new images or visual content, such as in graphic design or virtual reality. 

Examples of how image/visual generation drives data pipelines:

Creating virtual models for fashion designGenerating realistic game environments and charactersSimulating architectural visualizations for construction planningProducing visual content for marketing and advertisingDeveloping educational tools with custom illustrationsEnhancing film and video production with CGI effectsCreating personalized avatars for social media platforms

Recommender systems

Recommender systems are useful in a wide variety of applications, from e-commerce to content streaming services, where personalized suggestions improve user experience and engagement. 

Examples of how recommender systems drive data pipelines:

Personalizing content recommendations on streaming platformsSuggesting products to users on e-commerce sitesTailoring news feeds on social mediaRecommending music based on listening habitsSuggesting connections on professional networksCustomizing advertising to user preferencesProposing travel destinations and activities based on past behavior

Imagine this: It’s 2025. The CEO of a mid-sized tech company, overwhelmed by the rapid changes in AI, realizes the company is missing out. Despite having the latest tools and software, there’s still a gap—a missing strategic vision to make it all work seamlessly.

That’s when they decide to hire a Chief AI Officer. Within a year, the company transforms. Customer satisfaction is up, operations are smoother, and new revenue streams have opened. The CAIO didn’t just bring AI; they brought a revolution.

Artificial intelligence has evolved from an experimental technology to a core business necessity, reshaping operations, decision-making, and customer experiences. As its influence grows, so does the need for specialized leadership.

Enter the Chief AI Officer (CAIO), a role dedicated to embedding AI into the organization’s DNA. But what exactly does this role bring to the table that other tech executives might not?

Why a Chief AI Officer?

In many companies, AI initiatives have traditionally been managed by IT departments or overseen by roles like the Chief Data Officer (CDO) or Chief Technology Officer (CTO).

However, as AI’s impact broadens, the demand for dedicated AI leadership becomes clearer. A CAIO does more than oversee implementation; they shape how AI integrates with the organization’s core functions and long-term objectives.

Several critical factors underscore the rise of this role:

Specialized expertise in emerging AI applications: Implementing AI at a strategic level requires not only technical knowledge but also industry-specific insights. CAIOs need to stay ahead of AI’s evolving applications, including in non-traditional sectors like education, nonprofits, and disaster response. A CAIO with insights into these fields can tailor innovations to meet unique industry challenges, creating a distinct competitive advantage.Ethical and regulatory leadership: AI’s rapid adoption introduces pressing ethical and regulatory issues, from privacy concerns to managing bias. CAIOs play a crucial role in ensuring that AI systems adhere to ethical principles, such as those outlined in the UNESCO Recommendation on the Ethics of Artificial Intelligence. By establishing clear guidelines and monitoring AI’s impact, CAIOs can help mitigate potential harms, promote transparency, and foster public trust—elements critical for organizations that seek to lead responsibly in AI.Driving business transformation: The CAIO’s role goes beyond introducing AI tools; it’s about transforming business processes, opening new revenue streams, and improving customer experience. For instance, the grant proposal tool I implemented reduced preparation time by over 30 hours per proposal, illustrating the kind of measurable impact that a CAIO can bring. Positioned at the executive level, the CAIO drives AI initiatives that create significant, lasting change.Workforce development and transformation: The demand for AI talent is high, and a CAIO is essential in attracting, developing, and retaining team members who can deliver on AI strategies. They foster an AI-savvy culture that integrates technical and business knowledge across the workforce. By prioritizing internal training and upskilling, CAIOs can help employees embrace AI as a valuable tool, not a threat.Cross-departmental integration: AI’s reach extends to every corner of a business, impacting marketing, customer service, HR, and beyond. A CAIO ensures that AI adoption is cohesive and strategic, breaking down departmental silos to drive alignment with the company’s goals. For example, implementing an AI recommendation engine across product development and customer service can streamline and enhance the entire customer journey, delivering value at every touchpoint.

Key responsibilities of a Chief AI Officer

A CAIO’s responsibilities are diverse and strategic, encompassing the oversight of AI initiatives, risk management, and performance measurement. Key duties include:

Strategic planning: Develop a clear AI vision, prioritize high-impact projects, and collaborate with other executives to ensure AI initiatives align with organizational goals. Strategic planning with a CAIO is about more than timelines; it’s about identifying projects that will have meaningful, transformative impact.Implementation oversight: Oversee the end-to-end development and deployment of AI initiatives, ensuring each project—from model design to deployment—meets strategic objectives. CAIOs prioritize high-ROI projects and track their success to showcase AI’s tangible value within the organization.Governance and ethics: Establish ethical governance frameworks to manage biases, protect data privacy, and adhere to regulations, embedding responsible AI practices within the organization’s culture. In my work developing governance frameworks, I’ve built models to track and mitigate bias, highlighting that ethical AI governance is an ongoing process, not a one-time setup.Change management and education: Drive AI adoption across the organization by addressing concerns, promoting understanding, and providing upskilling opportunities. Educating employees about AI’s benefits is critical for fostering acceptance and creating a culture where AI is seen as empowering, not disruptive.Performance measurement and iteration: Set and monitor metrics—such as efficiency gains, revenue impact, and customer satisfaction improvements—to assess AI’s success. CAIOs continuously refine AI strategies to adapt to technological advancements, making performance measurement a cornerstone of AI leadership.

Is a CAIO right for your organization?

Not every organization may need a dedicated CAIO. For smaller businesses or those with limited AI applications, roles like the CTO or CDO might sufficiently cover AI needs.

However, companies with ambitious AI goals—especially in complex or regulated sectors like finance, healthcare, or retail—can gain substantial value from having a CAIO to focus on AI’s strategic alignment, ethical oversight, and cohesive deployment.

For organizations that aren’t yet ready to bring on a CAIO, developing CAIO-like responsibilities within existing roles can serve as a bridge. This approach prepares the organization to navigate AI’s growing influence, positioning it to embrace a future where the CAIO role might become essential.

The CAIO doesn’t just drive AI strategy; they align AI initiatives with the broader business vision, ensuring that implementations are impactful, ethical, and compliant. In an era where AI is integral to business success, a CAIO’s focused leadership could be the competitive edge that organizations need to stay ahead.

Conclusion

The emergence of the Chief AI Officer marks a pivotal shift in business, where AI becomes a strategic driver of innovation and a core element of corporate vision.

For organizations committed to responsible, comprehensive AI adoption, a CAIO can be the catalyst that unites people, processes, and technology, future-proofing the organization in an AI-powered world.

Transforming customer experiences, developing an AI-capable workforce, and establishing ethical standards, a Chief AI Officer (CAIO) plays a crucial role in driving the change needed to navigate today’s ever-evolving AI landscape.

Want more from Dr. Denise Turley?

Check out her other articles below:

سلام. خوش آمدید

In the very first episode of our monhtly Explainable AI podcas, hosts Paul Anthony Claxton and Rohan Hall sat down with James Chalmers, Chief Revenue Officer of Novo Power, to discuss one of the most pressing issues in AI today: energy consumption and its environmental impact.

Together, they explored how AI’s rapid expansion is placing significant demands on global power infrastructures and what leaders in the tech industry are doing to address this.

The conversation covered various important topics, from the unique power demands of generative AI models to potential solutions like neuromorphic computing and waste heat recapture. If you’re interested in how AI shapes business and global energy policies, this episode is a must-listen.

Why this conversation matters for the future of AI

The rise of AI, especially generative models, isn’t just advancing technology; it’s consuming power at an unprecedented rate. Understanding these impacts is crucial for AI enthusiasts who want to see AI development continue sustainably and ethically.

As James explains, AI’s current reliance on massive datasets and intensive computational power has given it the fastest-growing energy footprint of any technology in history. For those working in AI, understanding how to manage these demands can be a significant asset in building future-forward solutions.

Main takeaways

AI’s power consumption problem: Generative AI models, which require vast amounts of energy for training and generation, consume ten times more power than traditional search engines.Waste heat utilization: Nearly all power in data centers is lost as waste heat. Solutions like those at Novo Power are exploring how to recycle this energy.Neuromorphic computing: This emerging technology, inspired by human neural networks, promises more energy-efficient AI processing.Shift to responsible use: AI can help businesses address inefficiencies, but organizations need to integrate AI where it truly supports business goals rather than simply following trends.Educational imperative: For AI to reach its potential without causing environmental strain, a broader understanding of its capabilities, impacts, and sustainable use is essential.

Meet James Chalmers

James Chalmers is a seasoned executive and strategist with extensive international experience guiding ventures through fundraising, product development, commercialization, and growth.

As the Founder and Managing Partner at BaseCamp, he has reshaped traditional engagement models between startups, service providers, and investors, emphasizing a unique approach to creating long-term value through differentiation.

Rather than merely enhancing existing processes, James champions transformative strategies that set companies apart, strongly emphasizing sustainable development.

Numerous accolades validate his work, including recognition from Forbes and Inc. Magazine as a leader of one of the Fastest-Growing and Most Innovative Companies, as well as B Corporation’s Best for The World and MedTech World’s Best Consultancy Services.

He’s also a LinkedIn ‘Top Voice’ on Product Development, Entrepreneurship, and Sustainable Development, reflecting his ability to drive substantial and sustainable growth through innovation and sound business fundamentals.

At BaseCamp, James applies his executive expertise to provide hands-on advisory services in fundraising, product development, commercialization, and executive strategy.

His commitment extends beyond addressing immediate business challenges; he prioritizes building competency and capacity within each startup he advises. Focused on sustainability, his work is dedicated to supporting companies that address one or more of the United Nations’ 17 Sustainable Development Goals through AI, DeepTech, or Platform Technologies.

About the hosts:

In the latest episode of The Generative AI Podcast, host Arsenii Shatokhin sat down with Karanveer Anand, a Technical Program Manager at Google, to explore how AI is reshaping the field of program management.

They dove into everything from the role of AI in cloud computing to the evolving balance between AI innovation and safety. If you’re curious about how AI is influencing the future of technical program management, this discussion is a must-listen.

Catch the full episode right here.

Why AI enthusiasts should care

AI is no longer just a buzzword; it’s becoming a critical part of how businesses operate, making it essential for technical program managers to understand its implications.

As Karanveer explains, AI is revolutionizing how program management is done, introducing new efficiencies and ways to optimize workflows. For those working in AI and tech, understanding the intersection of AI and program management can provide a significant competitive edge.

Main takeaways

AI in program management: AI requires deep technical understanding and brings new challenges to program management.AI’s role in cloud computing: AI optimizes cloud resource allocation for better efficiency and cost management.AI as a necessity: Tools like Gemini integrate AI into everyday tasks, making it indispensable.Prioritizing AI safety: Safety should be built into AI frameworks from the start.AI’s future across industries: AI is set to transform sectors like healthcare and finance, offering new opportunities for innovation.

Meet Karanveer Anand

Karanveer Anand is a Technical Program Manager at Google, specializing in software reliability. His role focuses on ensuring that Google’s services remain highly reliable and accessible to billions of users worldwide.

With a background in cloud infrastructure and AI, Karanveer is passionate about using AI to improve software reliability and streamline processes. He is always exploring innovative ways to apply AI and is a thought leader in the intersection of AI, cloud computing, and program management.

As businesses seek to leverage AI to enhance their offerings and streamline operations, the ability to effectively implement AI solutions is critical to achieving sustained success. 

AI and deep learning inference demand powerful AI accelerators, but are you truly maximizing yours?

GPUs often operate at a mere 30-40% utilization, squandering valuable silicon, budget, and energy.

In this live session, NeuReality’s Field CTO, Iddo Kadim, tackles the critical challenge of maximizing AI accelerator capability. Whether you build, borrow, or buy AI acceleration – this is a must-attend.

Date: Thursday, December 5
Time: 10 AM PST | 5 PM GMT
Location: Online

Iddo will reveal a multi-faceted approach encompassing intelligent software, optimized APIs, and efficient AI inference instructions to unlock benchmark-shattering performance for ANY AI accelerator.

The result?

You’ll get more from the GPUs buy, rather than buying more GPUs to make up for the limitations of today’s CPU and NIC-reliant inference architectures. And, you’ll likely achieve superior system performance within your current energy and cost constraints. 

Your key takeaways:

The urgency of GPU optimization: Is mediocre utilization hindering your AI initiatives? Discover new approaches to achieve 100% utilization with superior performance per dollar and per watt leading to greater energy efficiency.Factors impacting utilization: Master the key metrics that influence GPU utilization: compute usage, memory usage, and memory bandwidth.Beyond hardware: Harness the power of intelligent software and APIs. Optimize AI data pre-processing, compute graphs, and workload routing to maximize your AI accelerator (XPU, ASIC, FPGA) investments.Smart options to explore: Uncover the root causes of underutilized AI accelerators and explore modern solutions to remedy them. You’ll get a summary of recent LLM real-world performance results – made possible by pairing NeuReality’s NR1 server-on-a-chip with any GPU or AI accelerator.

You spent a fortune on your GPUs – don’t let them sit idle for any amount of time.

Did you know that 27.1% of AI practitioners and 32.5% of AI tools’ end users don’t specifically address artificial intelligence’s biases and challenges? The technology is helping to improve industries like healthcare, where diagnoses can be improved through rapidly evolving technology. 

However, this raises ethical concerns about the potential for AI systems to be biased, threaten human rights, contribute to climate change, and more. In our Generative AI 2024 report, we set out to understand how businesses address these ethical AI issues by surveying practitioners and end users.

With the global AI market size forecast to be US$1.8tn by 2030 and AI being deeply intertwined with our lives, it’s vital to address potential issues. Ethical AI is developing and deploying systems to highlight accountability, transparency, and fairness for human values. 

Understanding AI bias

Bias can occur throughout the various stages of the AI pipeline, and one of the primary sources of this bias is data collection. Outputs are more likely to be biased if the data collected to train AI algorithms isn’t diverse or representative of minorities.

It’s also important to recognize other stages where bias can occur unconsciously, such as: 

Data labeling. Annotators can have different interpretations of the same labels.Model training. The data collected must be balanced, and the model architecture capable of handling diverse inputs must be balanced, or the outputs could be biased.Model deployment. The AI systems must be monitored and tested for bias before deployment.

As we increasingly utilize AI in society, there have been situations where bias has surfaced. In healthcare, for example, computer-aided diagnosis (CAD) systems have been proven to provide lower accuracy results for black female patients when compared to white female patients.

With Midjourney, academic research found that, when asked, the technology generated images of people in specialized professions as men looking older and women looking younger, which reinforces gendered bias. 

A few organizations in the criminal justice system are using AI tools to predict areas where a high incidence of crime is likely. As these tools can often rely solely on historical arrest data, this can reinforce any existing patterns of racial profiling, leading to an excessive targeting of minority communities. 

Challenges in creating ethical AI

We’ve seen how bias can exist in AI, but that isn’t the only one it faces. AI can potentially improve business efficiency, but there are a few challenges to ensuring that the ethics of AI solutions are a key focus.

1. Security

AI can be susceptible to hacking; the Cybersecurity Infrastructure and Security Agency (CISA) mentions documented times when attacks have led to objects being hidden from security camera footage and autonomous vehicles acting poorly.

2. Misinformation

With the potential to cause severe reputational damage, it’s essential to curb the likelihood of AI tools spreading untrue facts by establishing proper steps when developing the technology. Misinformation can affect public opinion and spread the wrong information as if it is true.

3. Job displacement

AI can automate various work activities, freeing up valuable worker time. However, this could lead to job loss, with lower-wage workers needing to upskill or change careers. Creating ethical AI also includes making sure that tools complement jobs and not replace them.

4. Intellectual property

OpenAI had a lawsuit involving multiple famous writers who stated their platform, ChatGPT, illegally used their copyrighted work. The lawsuit claimed that AI exploits intellectual property, which can lead to authors being unable to make a living from their work.

5. Ethics and competition

With the constant need to innovate, companies may need to take more time to ensure their AI systems are designed to be ethically sound. Additionally, strong security measures must be in place to protect businesses and users.

Strategies to address AI bias

We wanted to know how practitioners and end users of AI tools addressed biases and challenges, as companies need to be aware of steps that need to be taken when using this technology.

1. Regular audits and assessments

44.1% of practitioners and 31.1% of end users stated they addressed bias by regular auditing and assessing. This often includes a comprehensive evaluation of AI system algorithms, where the first step is to understand where bias is more likely to occur.

Following this, it’s vital to examine for unconscious bias, such as disparities in how AI systems handle age, ethnicity, gender, and other factors. Recognizing these issues allows businesses to create and implement strategies to minimize and remove biases for improved fairness. This could be changing the training data for AI models or proposing new documentation.

2. Rely on tool providers’ ethical guidelines

According to UNESCO, there are ten core principles to make sure ethical AI has a human-centered approach:

Proportionality and do no harm. AI systems are to be used only when necessary, and risk assessments need to be done to avoid harmful outcomes from their use.Safety and security. Security and safety risks need to be avoided and addressed by AI actors.Right to privacy and data protection. Data protection frameworks need to be established alongside privacy.Multi-stakeholder and adaptive governance & collaboration. AI governance is essential; diverse stakeholders must participate, and companies must follow international law and national sovereignty regarding data use.Responsibility and accountability. Companies creating AI systems need to have mechanisms in place so these can be audited and traced.Transparency and explainability. AI systems need appropriate levels of explainability and transparency to ensure safety, security, and privacy.Human oversight and determination. AI systems can’t displace human accountability and responsibility.Sustainability. Assessments must be made to determine the impact AI systems have on sustainability. Awareness and literacy. It’s vital to ensure an open and accessible education for the public about AI and data.Fitness and non-discrimination. To ensure AI can benefit all, fairness, social justice, and non-discrimination must be promoted.

28.6% of end users and 22% of practitioners rely on AI tool providers to follow appropriate ethical guidelines, so it’s essential that AI systems have ethical AI in all stages of development and deployment of their technology.

3. Don’t specifically address

A substantial percentage of end users and practitioners, 32.5% and 27.1%, respectively, said they don’t specifically address biases when using AI tools. With this technology being widely used across various industries, not addressing concerns and challenges could lead to further issues.

In addition to data bias, privacy is a top concern; smart home software, for example, must have robust privacy settings to prevent hacking or tampering. Similarly, AI systems can often make decisions that have a profound impact—autonomous vehicles must keep everyone on the road safe, and ensuring that AI doesn’t make mistakes is essential.

Crafting better AI tools

When creating AI tools, it’s important to focus on all aspects—ethical AI is, perhaps, the most vital component, as it affects output and how various minorities, such as gender and ethnicity, may be treated in industries like healthcare and law.

Our Generative AI 2024 report offers a comprehensive overview of how practitioners and end users use AI tools and how the sentiment is on the ground. Trust is fundamental for AI technology, so make sure to get your copy to learn how much confidence users currently have.

Navigating the field of large language model operations (LLMOps) is more important than ever as businesses and technology sectors intensify utilizing these advanced tools. 

LLMOps is a niche technical domain and a fundamental aspect of modern artificial intelligence frameworks, influencing everything from model design to deployment. 

Whether you’re a seasoned data scientist, a machine learning engineer, or an IT professional, understanding the multifaceted landscape of LLMOps is essential for harnessing the full potential of large language models in today’s digital world. 

In this guide, we’ll cover:

What is LLMOps?How does LLMOps work?What are the benefits of LLMOps?LLMOps best practices

What is LLMOps?

Large language model operations, or LLMOps, are techniques, practices, and tools that are used in operating and managing LLMs throughout their entire lifecycle.

These operations comprise language model training, fine-tuning, monitoring, and deployment, as well as data preparation.  

What is the current LLMops landscape?

LLMs. What opened the way for LLMOps.Custom LLM stack. A wider array of tools that can fine-tune and implement proprietary solutions from open-source regulations.LLM-as-a-Service. The most popular way of delivering closed-based models, it offers LLMs as an API through its infrastructure.Prompt execution tools. By managing prompt templates and creating chain-like sequences of relevant prompts, they help to improve and optimize model output.Prompt engineering tech. Instead of the more expensive fine-tuning, these technologies allow for in-context learning, which doesn’t use sensitive data.Vector databases. These retrieve contextually relevant data for specific commands.

What are the key LLMOps components?

Architectural selection and design

Choosing the right model architecture. Involving data, domain, model performance, and computing resources.Personalizing models for tasks. Pre-trained models can be customized for lower costs and time efficiency. Hyperparameter optimization. This optimizes model performance as it finds the best combination. For example, you can use random search, grid research, and Bayesian optimization.Tweaking and preparation. Unsupervised pre-training and transfer learning lower training time and enhance model performance. Model assessment and benchmarking. It’s always good practice to benchmark models against industry standards. 

Data management

Organization, storing, and versioning data. The right database and storage solutions simplify data storage, retrieval, and modification during the LLM lifecycle.Data gathering and processing. As LLMs run on diverse, high-quality data, models might need data from various domains, sources, and languages. Data needs to be cleaned and pre-processed before being fed into LLMs.Data labeling and annotation. Supervised learning needs consistent and reliable labeled data; when domain-specific or complex instances need expert judgment, human-in-the-loop techniques are beneficial.Data privacy and control. Involves pseudonymization, anonymization techniques, data access control, model security considerations, and compliance with GDPR and CCPA.Data version control. LLM iteration and performance improvement are simpler with a clear data history; you’ll find errors early by versioning models and thoroughly testing them.

Deployment platforms and strategies

Model maintenance. Showcases issues like model drift and flaws.Optimizing scalability and performance. Models might need to be horizontally scaled with more instances or vertically scaled with additional resources within high-traffic settings.On-premises or cloud deployment. Cloud deployment is flexible, easy to use, and scalable, while on-premises deployment could improve data control and security. 

LLMOps vs. MLOps: What’s the difference?

Machine learning operations, or MLOps, are practices that simplify and automate machine learning workflows and deployments. MLOps are essential for releasing new machine learning models with both data and code changes at the same time.

There are a few key principles of MLOps:

1. Model governance

Managing all aspects of machine learning to increase efficiency, governance is vital to institute a structured process for reviewing, validating, and approving models before launch. This also includes considering ethical, fairness, and ethical concerns.

2. Version control

Tracking changes in machine learning assets allows you to copy results and roll back to older versions when needed. Code reviews are part of all machine learning training models and code, and each is versioned for ease of auditing and reproduction.

3. Continuous X

Tests and code deployments are run continuously across machine learning pipelines. Within MLOps, ‘continuous’ relates to four activities that happen simultaneously whenever anything is changed in the system:

Continuous integrationContinuous deliveryContinuous trainingContinuous monitoring 

4. Automation

Through automation, there can be consistency, repeatability, and scalability within machine learning pipelines. Factors like model training code changes, messaging, and application code changes can initiate automated model training and deployment.

MLOps have a few key benefits:

Improved productivity. Deployments can be standardized for speed by reusing machine learning models across various applications.Faster time to market. Model creation and deployment can be automated, resulting in faster go-to-market times and reduced operational costs.Efficient model deployment. Continuous delivery (CI/CD) pipelines limit model performance degradation and help to retain quality. 

LLMOps are MLOps with technology and process upgrades tuned to the individual needs of LLMs. LLMs change machine learning workflows and requirements in distinct ways:

1. Performance metrics

When evaluating LLMs, there are several different standard scoring and benchmarks to take into account, like recall-oriented understudy for gisting evaluation (ROUGE) and bilingual evaluation understudy (BLEU).

2. Cost savings

Hyperparameter tuning in LLMs is vital to cutting the computational power and cost needs of both inference and training. LLMs start with a foundational model before being fine-tuned with new data for domain-specific refinements, allowing them to deliver higher performance with fewer costs.

3. Human feedback

LLM operations are typically open-ended, meaning human feedback from end users is essential to evaluate performance. Having these feedback loops in KKMOps pipelines streamlines assessment and provides data for future fine-tuning cycles.

4. Prompt engineering

Models that follow instructions can use complicated prompts or instructions, which are important to receive consistent and correct responses from LLMs. Through prompt engineering, you can lower the risk of prompt hacking and model hallucination.

5. Transfer learning

LLM models start with a foundational model and are then fine-tuned with new data, allowing for cutting-edge performance for specific applications with fewer computational resources.

6. LLM pipelines

These pipelines integrate various LLM calls to other systems like web searches, allowing LLMs to conduct sophisticated activities like a knowledge base Q&A. LLM application development tends to focus on creating pipelines, not new ones. 

How does LLMOps work?

LLMOps involve a few important steps:

1.  Selection of foundation model

Foundation models, which are LLMs pre-trained on big datasets, are used for downstream operations. Training models from scratch can be very expensive and time-consuming; big companies often develop proprietary foundation models, which are larger and have better performance than open-source ones. They do, however, have more expensive APIs and lower adaptability.

Proprietary model vendors:

OpenAI (GPT-3, GPT-4)AI21 Labs (Jurassic-2)Anthropic (Claude)

Open-source models:

LLaMAStable DiffusionFlan-T5

2. Downstream task adaptation

After selecting the foundation model, you can use LLM APIs, which don’t always say what input leads to what output. It might take iterations to get the LLM API output you need, and LLMs can hallucinate if they don’t have the right data. Model A/B testing or LLM-specific evaluation is often used to test performance.

You can adapt foundation models to downstream activities:

Model assessmentPrompt engineeringUsing embeddingsFine-tuning pre-trained modelsUsing external data for contextual information

3. Model deployment and monitoring

LLM-powered apps must closely monitor API model changes, as LLM deployment can change significantly across different versions.

What are the benefits of LLMOps?

Scalability

You can achieve more streamlined management and scalability of data, which is vital when overseeing, managing, controlling, or monitoring thousands of models for continuous deployment, integration, and delivery.

LLMOps does this by enhancing model latency for more responsiveness in user experience. Model monitoring with a continuous integration, deployment, and delivery environment can simplify scalability.

LLM pipelines often encourage collaboration and reduce speed release cycles, being easy to reproduce and leading to better collaboration across data teams. This leads to reduced conflict and increased release speed.

LLMOps can manage large amounts of requests simultaneously, which is important in enterprise applications.

Efficiency

LLMOps allow for streamlined collaboration between machine learning engineers, data scientists, stakeholders, and DevOps – this leads to a more unified platform for knowledge sharing and communication, as well as model development and employment, which allows for faster delivery.

You can also cut down on computational costs by optimizing model training. This includes choosing suitable architectures and using model pruning and quantization techniques, for example.

With LLMOps, you can also access more suitable hardware resources like GPUs, allowing for efficient monitoring, fine-tuning, and resource usage optimization. Data management is also simplified, as LLMOps facilitate strong data management practices for high-quality dataset sourcing, cleaning, and usage in training.

With model performance able to be improved through high-quality and domain-relevant training data, LLMOps guarantees peak performance. Hyperparameters can also be improved, and DaraOps integration can ease a smooth data flow. 

You can also speed up iteration and feedback loops through task automation and fast experimentation. 

3. Risk reduction

Advanced, enterprise-grade LLMOps can be used to enhance privacy and security as they prioritize protecting sensitive information. 

With transparency and faster responses to regulatory requests, you’ll be able to comply with organization and industry policies much more easily.

Other LLMOps benefits

Data labeling and annotation GPU acceleration for REST API model endpointsPrompt analytics, logging, and testingModel inference and servingData preparationModel review and governance

LLMOps best practices

These practices are a set of guidelines to help you manage and deploy LLMs efficiently and effectively. They cover several aspects of the LLMOps life cycle:

Exploratory Data Analysis (EDA)

Involves iteratively sharing, exploring, and preparing data for the machine learning lifecycle in order to produce editable, repeatable, and shareable datasets, visualizations, and tables.

Be part of a community

Stay up-to-date with the latest practices and advancements by engaging with the open-source community.

Data management

Appropriate software that can handle large volumes of data allows for efficient data recovery throughout the LLM lifecycle. Making sure to track changes with versioning is essential for seamless transitions between versions. Data must also be protected with access controls and transit encryption.

Data deployment

Tailor pre-trained models to conduct specific tasks for a more cost-effective approach.

Continuous model maintenance and monitoring

Dedicated monitoring tools are able to detect drift in model performance. Real-world feedback for model outputs can also help to refine and re-train the models.

Ethical model development

Discovering, anticipating, and correcting biases within training model outputs to avoid distortion.

Privacy and compliance

Ensure that operations follow regulations like CCPA and GDPR by having regular compliance checks.

Model fine-tuning, monitoring, and training

A responsive user experience relies on optimized model latency. Having tracking mechanisms for both pipeline and model lineage helps efficient lifecycle management. Distributed training helps to manage vast amounts of data and parameters in LLMs.

Model security

Conduct regular security tests and audits, checking for vulnerabilities.

Prompt engineering

Make sure to set prompt templates correctly for reliable and accurate responses. This also minimizes the probability of prompt hacking and model hallucinations.

LLM pipelines or chains

You can link several LLM external system interactions or calls to allow for complex tasks.

Computational resource management

Specialized GPUs help with extensive calculations on large datasets, allowing for faster and more data-parallel operations.

Disaster redundancy and recovery

Ensure that data, models, and configurations are regularly backed up. Redundancy allows you to handle system failures without any impact on model availability. 

Propel your career in AI with access to 200+ hours of video content, a free in-person Summit ticket annually, a members-only network, and more.

Sign up for a Pro+ membership today and unlock your potential.

I’m part of NVIDIA’s strategy team, and my background is in data.

My last role was as a data lead at Trade Republic, and I decided to change gears a little bit, do an MBA, and join NVIDIA in strategy. 

I have experience working in data in the US and then working in Germany. One of the reasons I felt frustrated with working in data is that I felt like, in Germany specifically, data teams were not necessarily at the forefront and weren’t necessarily building data products or being taken seriously.

That’s part of why I decided to join NVIDIA; it resonated with me, as well as the mission of pushing AI forward in Europe.

The slow adoption of generative AI in Europe

Europe is too slow at adopting Gen AI. 

The reality is that Europe is way behind the US in implementing the technology. 

The reason I’m highlighting this is because experimenting now is extremely important. The technology is moving forward very fast. 

Last year, we were talking about LLMs. This year, we’re talking about agentic workflows. So, it’s important to start experimenting with the technology now. Otherwise, it might be too advanced for us to catch up, and business models might become irrelevant.

It’s a very low risk to experiment with the technology because there’s a lot of proven value out there. If you look at research, you see that. The technology brings a lot of productivity gains.

A research paper from McKinsey that came out two weeks ago talks about forty percent of developers’ and product managers’ time being saved.

We’re also talking about three to fifteen percent higher revenues and ten to twenty percent ROI for companies implementing Gen AI.

Evolving customer expectations and business models

Customer expectations will keep evolving, and this technology will change business models.

They’re already expecting a high level of personalization from companies that they buy from regularly. About seventy percent of customers have that expectation, which will keep evolving. 

I use Amazon when ordering online in Berlin, and I have such great customer service that it’s hard to use another application. It becomes an expectation that you get that level of service. It’s important to think about this technology and how it will affect your industry. And, like I said, we’re yet again behind our American counterparts.

Background

The United States of America (USA) is one of numerous major economies taking forward a program for Artificial Intelligence (AI) Regulation. To ensure the USA plays a leading role in Artificial Intelligence research and development, the National Artificial Intelligence Initiative Act of 2020 was introduced and became law in 2021.

The Act’s overarching aim, inter alia, was to provide a broader initiative within the United States to ensure academia, the public, and private sectors could monitor and evaluate the performance of AI-based systems both before and post-deployment [1], [2].

Following this in 2022, the White House Office of Science and Technology Policy introduced the Blueprint for an AI Bill of Rights [3]. Following a year of public engagement to inform the creation of the framework, it outlines five core principles and associated practices to guide the creation, management, and iteration of automated systems while ensuring the protection of the American public’s rights [3].

With OpenAI introducing ChatGPT (Chat Generative Pre-Trained Transformer) in November 2022 and its forecasted economic potential exceeding $2.1 trillion [4], concerns were raised about how the technology at its current and continuous speed could work within the parameters of the justice system.

With the increasing adoption of Generative AI, President Biden in October 2023 issued an executive order on safe, secure, and trustworthy Artificial intelligence, stipulating (in amongst other requirements) that those developing the most powerful AI systems share their safety test data with the US government. For example, any companies developing foundation models that risk national security, national economic security, or national public health are required to inform the federal government during model training in addition to sharing red-team safety tests. [5].

The role of US Antitrust Laws

AI is seeing rapid expansion across sectors and organizations of all shapes and sizes, and therefore, the interoperability between AI as a tool and existing antitrust laws has been and will continue to be tested. While some states continue to work towards localized AI regulation, some argue that the pace and advancement around AI require a rewrite of antitrust laws.

For context, back in 1890, Congress passed the Sherman Act: a charter whose aim was to preserve free and unrestrained competition. Then, in 1914 a further two antitrust laws were passed, namely the Federal Trade Commission (FTC) Act (which created the Federal Trade Commission) and the Clayton Act [6], each of which are still in effect to this day.

Generally speaking, Antitrust laws exist to prevent unlawful mergers and business practices, with judgment being put in the hands of the courts to determine which cases are illegal based on the facts of each case. For over a century, these laws have retained the same core principle: protect competition to benefit consumers through the operation of operational efficiencies, fair pricing, and high-quality goods and services.

In summary, the Sherman Act makes illegal “every contract, combination, or conspiracy in restraint of trade” along with any “monopolization, attempted monopolization or conspiracy or combination to monopolize” [6].

The Supreme Court, however, ruled a while back that only unreasonable acts are prohibited: not every restraint of trade is included. For example, a partnership agreement between two individuals may restrain trade but not unreasonably and thus may be lawful under US antitrust law.

Any acts, though considered harmful to competition, are almost always illegal, and are known as per se violations, which include arrangements between businesses to fix prices, divide markets, or rig bids.

The Sherman Act can be enforced both in civil and criminal law, and both businesses and individuals can be prosecuted under it by the Department of Justice.

If a competitor fixes prices or rigs bids, penalties can include up to $100 million for corporations and $1 million for an individual, along with ten years imprisonment. Under federal law, the maximum fine can be increased to twice the amount the conspirators gained from the illegal activity or, on the other hand, twice the money lost by the victims if either of these amounts exceeds $100 million [6].

The Clayton Act, however, addresses more specific areas the Sherman Act does not clearly prohibit. Section 7 of the Clayton Act prohibits mergers and acquisitions demonstrating anti-competitive effects, to quote, “may be substantially to lessen competition, or to tend to create a monopoly.”

A further amendment in 1976 of the Clayton Act by the Hart-Scott-Rodino Antitrust Improvements Act requires advance notice from organizations planning a large merger or acquisition: they must notify the government of their plans.

It is important to note that private parties under the Clayton Act authorize private parties to sue for triple damages if they have been harmed by conduct that is in violation of either the Sherman or Clayton Act. Additionally, they can obtain a court order prohibiting the anticompetitive practice in the future. [6]

USA approaches to AI regulation example: Colorado AI Act

Antitrust laws aside, states are taking differing approaches in trying to regulate AI. The Colorado AI Act (also referred to as Consumer Protections for Artificial Intelligence), for example, was signed into law on the 17th of May 2024 but does not come into effect until the 1st of February 2026 [7].

While there are similarities between it and the EU’s AI Act, the Colorado AI Act specifically focuses on high-risk AI systems. Developers are required to put in place safeguards by sharing information with deployers, such as what data has been used for model training, risk mitigation measures, and reasonably foreseeable limitations of the system.

Additionally, developers must publicly share information on their website or in a public use case inventory two key pieces of information: 1) the type of high-risk system they developed and 2) what steps they are taking to manage risks of algorithmic discrimination. Most importantly, should algorithmic discrimination occur through the intended use of the system, developers must disclose this to both the Colorado Attorney General and the system developers in question. 

Alongside developers, deployers must follow a risk management policy such as the National Institute of Standards and Technology’s Artificial Intelligence Risk Management Framework and the International Organization for the Standardizations Standard ISO/IEC 420001.

Among these existing standards, the size and complexity of the deployer will need to be factored into the reasonability of the framework. Should the system undergo an intentional or substantial modification, deployers must conduct an impact assessment within 90 days of the modification taking place in addition to two additional impact assessments for the system, specifically an impact assessment for the system alongside an annual impact assessment for any system deployed. 

Similar to developers, deployers will need to publish information on their website and disclose any occurrences of algorithmic discrimination to the Colorado Attorney General. There are some cases, though, in which exemptions can be granted: if deployers have an employee headcount under 50, they can be exempt from most of the requirements, providing certain conditions have been satisfied. [7]

Does the USA have a long road ahead behind the AI Act?

In September 2024, the European Commission’s EU Competitiveness Report highlighted that 30% of unicorn startups founded in Europe between 2008 and 2021 were relocating abroad, with many of them to the USA [8].

It will therefore be imperative that while the technology conglomerates continue to innovate in the AI space, lead by example, and collaborate closely with the US government, safeguards for fundamental rights and product safety must be in place but not be excessively restrictive to the point they prevent smaller players from the development and adoption of frontier AI.

When it comes to digital competition, the European Union’s Digital Markets Act and Digital Services Act ensure fair online market practices are enforced. In contrast, in the US there are no digital-specific competition laws. However, two pending pieces of legislation, namely the American Innovation and Choice Online Act (“AICOA”) and the Open App Markets Act (“OAMA”), could, if passed, result in drastic changes to American regulation of digital competition with the aim of targeting companies such as Google, Apple, Meta, Amazon, Microsoft and possibly TikTok [9].

A multilateral approach to managing, understanding, and implementing AI regulation will be required to in the long run assess whether laws around AI technologies can be fairly but rigorously enforced.

The recent executive order, the introduction of state-level AI-specific laws, and the voluntary commitment from influential AI companies (i.e., OpenAI, Meta, and Google) to increase testing of AI systems alongside sharing information on managing AI risks are important steps in understanding this fast-paced technology.

It will not, however, change the challenge of a lack of a singular definition of AI, and instead, it could be seen as a hard yard ahead in identifying that any material shift in antitrust regulation more closely aligned to AI innovation may only be feasible if regulating the outcome of AI becomes the focus instead of the attempt to holistically regulate AI. 

Bibliography

[1] Parker Lynne, Director of the National AI Initiative Office, Deputy United States Chief Technology Office, ‘National Artificial Intelligence Initiative’ (Artificial Intelligence and Emerging Technology Inaugural Stakeholder Meeting, June 29, 2022) < www.uspto.gov/sites/default/files/documents/National-Artificial-Intelligence Initiative-Overview.pdf > accessed 1st October 2024. (OSCOLA)

[2] H.R.6216 – 116th Congress (2019 – 2020): National Artificial Intelligence Initiative Act of 2020, 116th Cong. (2020), https://www.congress.gov/bill/116th-congress/ house-bill/6216 (BlueBook – change to OSCOLA)

[3] ’Blueprint for an AI Bill of Rights’ (Office of Science and Technology Policy, The White House) < www.whitehouse.gov/ostp/ai-bill-of-rights/ > accessed 1st October 2024. (OSCOLA)

[4] ’Economic Potential of Generative AI: The Next Productivity Frontier’ (McKinsey Digital, 14 June 2023 ) < www.mckinsey.com/capabilities/mckinsey-digital/our insights/th e-economic-potential-of-generative-ai-the-next-productivity-frontier#/> accessed 3rd October 2024.

[5] ’President Biden Issues Executive Order on Safe, Secure and Trustworthy Artificial Intelligence’ (The White House, Briefing Room Statements and Releases) < www.whitehouse.gov/briefing-room/statements-releases/2023/10/30/fact-sheet president-biden-issues-executive-order-on-safe-secure-and-trustworthy-artificial intelligence/ > accessed 1st October 2024.

[6] ’The Antitrust Laws’ (Federal Trade Commission – Competition Guidance) < www.ftc.gov/advice-guidance/competition-guidance/guide-antitrust-laws/antitrust laws> accessed 30 September 2024.

[7] ‘Colorado Governor Signs Comprehensive AI Bill’ (Mayer Brown, Insights) <www.mayerbrown.com/en/insights/publications/2024/06/colorado-governor-signs comprehensive-ai-bill > accessed 1st October 2024.

[8] ’The future of European Competitiveness: Part B’ (European Commission) <https://commission.europa.eu/document/download/ec1409c1- 

d4b4-4882-8bdd-3519f86bbb92_en?filename=The future of European competitiveness_ In-depth analysis and recommendations_0.pdf > accessed 1st October 2024.

[9] B Hoffman, ‘Digital Markets Regulation Handbook’ (Cleary Gottlieb, January 2024)

< https://content.clearygottlieb.com/antitrust/digital-markets-regulation-handbook/ united-states/index.html > accessed 3rd October 2024.

Looking for 200+ hours of expert AI advice?

Our Pro+ membership gives you access to videos of all of our past events, plus frameworks and templates.

But there’s more…

Sign up today and unlock your full potential.

Few companies have made as significant an impact on the creative sectors as Runway has. As a beacon of innovation in generative AI, Runway has carved out a reputation for transforming art, entertainment, and human creativity. 

At the helm of these groundbreaking endeavors is Anastasis Germanidis, the Co-Founder and Chief Technology Officer of Runway, who has played a pivotal role in shaping the future of creative technologies.

Recently recognized by Time Magazine as one of the “100 most influential companies” in 2023, Runway continues to be at the forefront of the AI revolution, creating tools that enhance the capabilities of creatives worldwide and redefining what is possible in storytelling and artistic expression. 

In this exclusive interview, Anastasis shares his journey from the inception of Runway at NYU’s ITP program, alongside co-founders Cristobal and Alejandro, to the cutting-edge developments that continue to push the boundaries of computational creativity.

Psst. Anastasis will be at our summit in Boston.

Why not get your tickets and attend his talk?

Please introduce yourself and briefly introduce your journey leading up to this point with Runway.

I met my two co-founders, Cristobal and Alejandro, at NYU’s ITP program in 2016. Our shared curiosity about the potential of computational creativity led us to build tools for our peers—we wanted to build tools that let them interact with what was, at the time, an emerging technology, and that’s ultimately how Runway was originally born. 

Today, we are a full-stack, applied AI research company that builds generative AI systems and tools for creatives of all backgrounds.

Runway has become a pioneering platform in creative AI tools. What was the company’s original vision, and how has that evolved?

Being artists ourselves, we started Runway as a company for artists by artists. That vision has been central to our ethos since the very beginning and remains true today.

As our research and capabilities have continued to advance over the last few years – most recently with the release of our newest foundational model, Gen-3 Alpha – our original vision has remained the same: we’re enabling new ways of bringing stories to life and opening doors for new storytellers.

Runway’s tools empower creatives to work with AI in previously unimaginable ways. What are the most exciting use cases that have emerged from your platform?

Everyone, from Fortune 500 and Global 2000 companies to freelancers, marketers, and Hollywood studios, uses our tools to tell new types of stories and streamline workflows.

The generative AI landscape is growing rapidly. In your opinion, what are the biggest challenges and opportunities facing the industry today?

When we started Runway back in 2018, we were some of the only ones building in this space, so it’s been incredible to see the advancements we’ve made as an industry in the last couple of years and to see the creativity that these tools have unlocked for artists. 

That said, we’re still very early in the lifecycle of these tools, and there is a lot still to be built and unlocked, especially when it comes to further improving quality and introducing new control mechanisms.

Something our team is currently focused on that will continue to unlock creativity is the development of General World Models – these are systems that understand the visual world and its dynamics. Gen-3 Alpha has been a major step toward this goal, but it’s still very early.

How do you see the relationship between human creativity and AI evolving over the next few years? Will AI enhance human creativity, or do you foresee it changing creative industries in more fundamental ways?

The history of art has always been deeply intertwined with the history of technology. As our technical capabilities continue to expand, the tools will continue to expand, but they’ll always be in the service of human artists and creators.

Runway is at the forefront of generative AI for creators. What innovations or features can users expect shortly that will further transform their workflows?

Gen-3 Alpha is the first and smallest of upcoming models and a major step toward building general world models, but there’s still more work to be done. For example, the model can struggle with complex character and object interactions, and generations don’t always follow the laws of physics precisely.

General world models will aim to represent and simulate a wide range of situations and interactions, like those encountered in the real world, and we’ll continue to build towards that future.

Looking ahead, what excites you the most about the future of generative AI, both in terms of its potential for creativity and broader applications?

Generative AI is still incredibly young, and we’re discovering new use cases daily. We recently announced a partnership with Lionsgate Studios, marking a significant milestone in the collaboration between AI and Hollywood and unlocking new opportunities to evolve workflows and offer brand-new tools to the entertainment industry.

Want more from Anastasis?

He’ll be at our summit in Boston this month on October 17.

His talk on ‘Shaping the next era of art, entertainment and human creativity’ will be invaluable.

Get your tickets today.

Sustainability vs scalability

Artificial Intelligence (AI) is the most disruptive technology of our time, but as its impact continues to unfold, many are questioning whether it’s just another tech fad, or if it’s here to reshape the future permanently.

The rapid rise of AI has been met with both excitement to revolutionize our world and skepticism and growing concern about the long-term feasibility and ethical risks.

So, are we being duped into believing AI is scalable enough to solve all of our problems? Or will it stand the test of time, proving itself sustainable to humanity and our evolving world? These are the questions I want to address in this writing.

Ahh the question of “Is AI sustainable? No, it is not sustainable, it is scalable, and that is what will make it sustainable.

I think digging into whether AI is sustainable is quite simply how we approach it. AI is a lot like engineered transportation. Vehicle mobility has taken us from being foot mobile or horse and buggy to traveling around the world in just a matter of hours.

This is what AI has and will continue to do for us. Just like engineered transportation has not gone anywhere, nor will it ever, nor will AI. The problems that exist with engineered mobility over the last 135-plus years still exist today by and large.

For example, cars still wreak havoc on our environment, and they are dangerous at points depending on how they are controlled, but that does not mean they have not been scalable or become obsolete. It is clear for humans the benefits of engineered mobility have far outweighed the collateral damage.

We have worked hard and continue to work hard to put laws and regulations in place at the organizational and consumer levels as to how cars and even planes and the like are built, used and operated. But cars are definitely not sustainable, if they were then we wouldn’t junk them every 10 years.

We cannot continue to operate engineered mobility the way we do because it does have its dangers, and it does wreak havoc on our earth. AI is much the same and it is not sustainable in its current existence. 

Scalability and sustainability are closely connected, but they are distinct concepts that often get confused very often. 

Most inventions start off in a phase where they are not inherently sustainable. At this early stage, they might be too costly, resource-intensive, or inefficient to sustain long-term.

However, as they scale—reaching larger markets, benefiting from economies of scale, improving their technology, and optimizing their production processes—they can become more efficient, less costly, and more adaptable to different environments. This scaling process is what ultimately feeds into making them sustainable.

The definition of sustainability

I think we would all agree that we cannot maintain engineered mobility in the way that we have in the past in an ongoing manner. We also cannot continue to deplete our resources and disrupt the ecological balance. It is not feasibly sustainable. It does not mean engineered mobility is going anywhere, rather, it is going to scale, which means it is going to get better, and eventually, sustainable. 

Sustainable efforts in transportation (e.g., electric vehicles, fuel efficiency, regulations) have aimed to improve this. In the same way, AI sustainability focuses on minimizing energy consumption, reducing environmental impact, and ensuring fair use of resources and more energy-efficient models, similar to how transportation is moving towards electric vehicles.

I think people tend to get scalability and sustainability mixed up. Scalability feeds sustainability. There is a difference between sustainability and scalability. Engineered mobility has scaled continuously over the last century and a half, but it has never reached a point of full sustainability.

AI is similar.

The problems and challenges we face in AI does not mean AI is going anywhere, it just means it is going to get better, and eventually, sustainable.

The definition of scalability

The parallel drawn between AI and engineered mobility makes it clear that when first invented, both of these inventions were expensive, inefficient, sometimes unsafe and available to only a select few.

Over time, the scaling of manufacturing processes, infrastructure development, and technological improvements made them more affordable, more reliable, and widely accessible, thereby achieving better sustainability in their use and production.

Like engineered mobility, but in the context of AI, we see that it is currently highly scalable—new models can be deployed, adjusted, and integrated across various industries and applications with relative ease. 

I think this is what we can expect for the future of AI. 

To be or not to be, that is the question

“To be, or not to be, that is the question: Whether ’tis nobler in mind to suffer The slings and arrows of outrageous fortune, Or to take arms against a sea of troubles And by opposing end them.”

— William Shakespeare, Hamlet

To be

Some early adopters swear by AI’s potential to revolutionize industries, while others remain concerned about the long-term feasibility, sustainability, and ethical implications.

On the first account, the power of AI is already showing its muscle in areas like healthcare, finance, marketing, and many other sectors. With billions of dollars backing AI behind major corporations and high-profile investment entities, it is clear people are serious about AI and its promising future.

Or not to be

On the other hand ongoing issues with data privacy concerns, potential job displacement, power consumption, data shortages, and the challenge of creating explainable AI models leads many critics to believe we could all just be “duped” because we have adopted AI technologies prematurely without fully understanding their implications or long-term viability. 

All of this skepticism extends to the possibility that as soon as AI could lose momentum, leading to a mass exodus of interest or what could be referred to as AI’s “dumping point.”

In reality, the true future of AI likely lies somewhere in between. Some applications are failing to scale and live up to the hype—similar to the many failed “AI wrapper” companies associated with the financial struggles of 2023 which began with the Silicon Valley Banking Collapse—while others are evolving to become much more integral to the way we live and work.

We are all AI bullish

I want to start with a question that someone in the business community said to me, “You seem so bullish on AI.  I feel like we are in a bubble, and when we emerge, we will feel duped. But you feel so bullish – am I wrong?” I responded, “I am bullish on AI, and so are you. If you took away some of the things that make your life livable today, you’d quickly realize just how bullish you are. You’re probably using AI without even knowing it. Yes, I agree we are in a bubble because we are using AI many times when we do not know or fully understand its implications and consequences, which can lead to the overconfidence that produces market bubbles.”

How are we bullish? Well imagine having to do without some of your creature comforts today. When we start explaining to the very people who are reticent about AI the things which makes their lives comfortable and maintainable that they don’t typically think about are ran, powered and enabled by AI allows them to a paradigm shift to actually begin to realize how bullish they are on AI. 

It’s like a person who would never eat a particular food product, but the end food product contains the ingredient of the particular food product they hate, and they love the end food product. 

For example, Castoreum is used in Vanilla Flavoring, and Castoreum is a secretion from the anal glands of beavers. Something that would send most people praying to the toilet gods. Castoreum is still approved by the FDA as a “natural flavoring,” and it has been used in everything from Vanilla Ice Cream, Chewing Gum, to Alcoholic Beverages.

We can live without AI, but you wouldn’t want to

Living without AI would be a nightmare—a chaotic, life-threatening plunge into inefficiency and disaster. Every moment of your life would be slower, more stressful, and infinitely more dangerous.

Imagine scrambling through endless paperwork, wasting hours on menial tasks that AI currently handles in seconds, and facing constant delays in every transaction, whether you’re trying to get healthcare, travel, or even buy food. Businesses would spiral into inefficiency, grinding to a halt, leaving the economy in tatters and unemployment through the roof.

Healthcare systems would be crippled, with life-or-death decisions left to outdated manual processes, causing countless avoidable deaths and suffering as diagnoses are delayed, treatments are wrong, and medical errors skyrocket. Important infrastructure like power grids and supply chains would falter, causing massive blackouts, food shortages, and transportation collapses that could spark widespread panic and societal breakdown.

Security would be practically nonexistent; criminals would have a field day with weakened defenses, leading to rampant crime, data breaches, and potentially catastrophic cyber-attacks. AI has become our silent guardian, the shield that holds back chaos—and without it, the world would rapidly descend into a hellish, volatile, and dangerous state that would make survival, let alone daily living, a grueling ordeal.

Removing AI would disrupt productivity, critical systems, and innovation, drastically impacting how society functions and progresses. Removing AI in many ways would be like putting the entire world back several decades. We would potentially have massive blackouts, disease, crime, and very possibly large-scale conflicts. Removing AI would make it difficult, almost impossible, to revert to a world without it without facing significant consequences.

AI has now become integral in modern life, deeply cemented in daily routines, critical infrastructure, and industry workflows, making it nearly impossible to revert to a world without it. It drives efficiency, automates complex tasks, and provides data-driven insights that shape decision-making across sectors. 

What are some examples of the things that are empowered by AI that we could not live without? 

Navigation and ride-sharing services are closely connected, as they rely heavily on AI-enhanced satellite and GPS positioning systems to optimize routes, predict traffic, and provide estimated arrival times. Without the AI processing of GPS data, ride-sharing technologies like Uber and Lyft would struggle to operate efficiently.

Similarly, energy and utility management benefit from smart grid infrastructure powered by AI, which helps manage energy distribution more effectively and predict power demand spikes, ultimately preventing power outages. Without AI these things would not operate correctly. 

A real-life incident: To live without AI is not living at all, literally

In 2016, a Tesla Model S, while in Autopilot mode, was traveling on a highway in the Netherlands and the vehicle’s AI-powered system detected that the car in front was about to collide with another car and automatically applied the brakes before the human driver could react.

The AI system used radar, cameras, and sensors to not only detect the car directly ahead but also analyze the traffic situation beyond it. The Autopilot’s collision avoidance system recognized that the vehicle ahead was decelerating suddenly and predicted an imminent crash.

Before the human driver had time to react, the Tesla autonomously braked, avoiding what could have been a high-speed collision that might have caused severe injuries or even death.

In the end, the AI saved both the driver and passengers from a life-threatening accident by reacting faster than a human could have in a dangerous situation. 

Without AI the driver would have potentially died, and so in this case, to live without AI is not living at all.

Have we been duped by AI?

Many of us use AI frequently without realizing it, which can make us overconfident and careless, like bulls in a china shop… 

In a way, by charging ahead without fully understanding AI, we risk becoming uninformed and ignorant about its capabilities and limitations, which leads to hard lessons. This can leave us vulnerable to deception and making decisions without all the information, leading us to feel like we have been duped.

This is a form of Black Box AI. What is Black Box AI and how can it be avoided?

Black box AI refers to artificial intelligence systems whose processes are opaque and difficult to interpret and fail to provide clear explanations for how they arrived at those outcomes. This lack of transparency can lead us to a feeling of being “duped”.  

As an everyday user, if you are unfamiliar with AI, the best way to protect yourself is to be cautious and informed about the AI tools you use. Start by choosing reputable apps and platforms with good privacy policies and transparent explanations of how they use AI.

If you are asked to provide personal data, make sure you understand why it’s needed and how it will be used. Pay attention to any unexpected behaviors or decisions made by AI systems, and don’t hesitate to question or report issues if something seems off.

It’s also helpful to regularly review the permissions you grant to apps and to be cautious about sharing sensitive information. If possible, use platforms that allow you to adjust settings, like opting out of data collection or personalization features. Being mindful of these basics will help you stay informed about what and how you are using AI and ensure you do not become a victim of AI or a perpetrator of AI misuse. 

The illusion of deception It’s easy to think that AI is deceiving us when things go wrong. We see rogue AI systems producing unexpected or incorrect outcomes, and it feels like we’ve been led astray. However, AI isn’t inherently deceptive; it doesn’t have intent or motive. Rather, it operates within the constraints and biases of the data and instructions it’s been given.

So I would say, no we have not been duped by AI, if we have been duped it is likely entirely our fault because of our approach to adopting AI. The responsibility relies on how humans engage with and understand AI, rather than suggesting that AI itself is inherently duping us.

As AI continues to evolve and take on more complex roles in society, from autonomous vehicles to healthcare and financial markets, it’s important that we educate ourselves about how these systems work so we don’t get duped.

Will we be dumped by AI

Will we, as humans, eventually be “dumped” by AI? In other words, will AI take over our jobs, outpace us intellectually, or leave us behind, creating a future where human roles are diminished or obsolete?

Myth: I’m just a mere human, I am no match for AI

One of the most common concerns about AI is the fear of job displacement. In industries like manufacturing, repetitive tasks have been automated by machines for years, but AI is now capable of handling more complex roles that involve sophisticated, yet contained decision-making, such as data analysis, and customer interaction. 

AI makes the honor roll

Another concern is with the development of increasingly sophisticated machine learning models, the fear is that AI will eventually outthink us, leaving humans intellectually inferior and eliminating your kids from making the honor roll.

The fear that AI will render humans “intellectually inferior” and eliminate opportunities for jobs and achievements like making the honor roll is exaggerated. AI can process information quickly and make complex decisions, but human qualities like creativity, emotional intelligence, critical thinking, and adaptability are currently beyond AI’s reach.

The job market and educational systems are going to adapt, focusing on skills that complement AI rather than compete with it, therefore enhancing human intelligence and capability.

This is a big opportunity to drill down on the creativity side and team up with AI’s intellectual prowess for those who feel AI is a displacement. 

AI operates as sophisticated yet contained intelligence. What does that mean? It means that AI operates within parameters, not footloose and voyaging as humans operate, but rather as an aid to humans and our creativity.

This leaves humans to evolve and expand into new realms where creativity can take us and show us why, but we can then bring AI into the picture, and it can show us how.

The beauty of this is that it allows for very accelerated learning and innovation to happen and for the global collective of humans to put their energy towards creativity and all of the productive ways that AI can be used and deployed. Imagine any idea you’ve ever had—what if you had access to all the resources needed to make it happen? 

Think about those questions or ideas you’ve searched everywhere to find answers for; this is where large language models like ChatGPT come in. They take your questions or ideas and draw upon a large range of relevant resources available, sometimes all resources available, to show you the way.

Now, imagine leveraging this capability for every resource on Earth. In the future, AI may enable us to tap into all imaginable resources to tackle even the most ambitious challenges, such as traveling to distant places in the universe light years away, far beyond the physical traveling capability of any human right now. This is the future of AI. This is the reality.

I don’t know if this is working; it just feels like we are two separate people

There is a fear that a type of “AI divide” could lead to economic and social disparities, with only those who have the resources to leverage AI succeeding, while others are left in less economically viable positions.

While the history of economics and human behavior shows that this outcome is somewhat likely, it is more probable that the AI divide will begin to evaporate as humans and machines work together to improve infrastructure, education, and cultural adoption.

Over the coming years and decades, as AI becomes increasingly consumerized, shared access and AI infrastructure will expand at the individual level, creating opportunities for everyone who desires access. Another reason the AI divide may lose its grip is that division often arises from a lack of opportunity, objectivity, and understanding.

AI has the potential to enable humans to operate at their best capacity—creativity and innovation—where opportunity thrives. As an objective technology, AI can help reduce the subjectivity in human decision-making that has plagued and divided societies since our existence.

On the other hand, AI could inadvertently “dump” portions of humanity by reinforcing biases, displacing workers, or concentrating power in the hands of a few. To support this, governments, organizations, and industries must collaborate to establish frameworks and policies that promote inclusive access to AI technologies. This means investing in education and retraining programs that equip workers with the skills needed for an AI-driven economy.

Take control of the relationship

The key to avoiding being “dumped” by AI lies in how we, as individuals and societies, adapt to the changes AI brings. It’s important to approach AI not as a threat but as a tool that can be leveraged for human advancement. 

In the end, the most important question isn’t whether AI will “dump” us, but rather, will we allow it to?

AI is here to stay, but its future lies in a balance between scalability and sustainability. As disruptive as it is, AI’s rapid growth has created both opportunities and challenges, leaving us to question if we’ve been prematurely optimistic or overly cautious about its potential.

Much like engineered mobility, AI is not yet sustainable, but it is scalable—and scalability is the path to sustainability. From revolutionizing industries to empowering daily life, AI’s current scalability drives efficiency and innovation, setting the stage for eventual sustainability as technology, regulation, and ethical considerations catch up.

While skepticism and challenges remain, the world is inevitably moving forward with AI, shaping a future where its integration is not just advantageous but very important for growth and survival. 

Human agency is central to the development and application of AI technologies. As creators and users of AI, we have the ability to steer its development in ways that benefit society rather than displace or harm it.

Through responsible innovation, thoughtful policy, and ongoing education, we can ensure that AI continues to serve as a powerful tool for human progress rather than a force that leaves us behind. Whether you believe we are being duped or on the brink of a revolutionary, humanity-changing evolution, one thing is certain: AI is an unstoppable force that will continue to reshape the world for generations to come.

Want access to hundreds of hours from our events?

Sign up for our membership and start watching today:

My name is Christie Mealo, and I’m a Senior AI Engineering Manager at CVS Health, where I focus on AI-driven health products, primarily in the area of diabetes management. 

In addition to my work at CVS, I’m the founder of Orbit, an AI-powered contact book and networking app designed for value-based networking. 

I also lead the Philly Data & AI Meetup group, help guide the Philly Tech Committee, and serve as a chair on Philly iConnect. 

Through these roles, I’m deeply involved in organizing communities and events across Philadelphia and the larger East Coast, helping to foster collaboration and innovation in the tech space.

It’s been a crazy year for those in tech—what’s excited you most about recent developments?

It’s been an incredible year in tech, and what excites me most is how generative AI has significantly lowered barriers to entry and creativity for so many people. This technology is empowering individuals with new and novel ideas, allowing them to bring their visions to life in ways that were previously out of reach. 

I believe this will shake up the economy in a positive way, leading to the development of a lot of innovative products and introducing new competitors into the market. While we’re undoubtedly in the midst of a hype cycle—or perhaps only at the beginning—it’s thrilling to see where this will take us in the coming years.

What role do you see generative AI playing across industries over the next 6-12 months, and where do you think it will have the biggest impact?

Generative AI is poised to significantly impact various industries over the next 6-12 months. While it’s clear that it will continue to transform fields like copywriting, advertising, and creative content, its influence is much broader.

On one hand, generative AI is incredibly exciting because it lowers barriers to entry for innovation and creativity. Tools like ChatGPT, Claude, Gemini, and GitHub Copilot are not only enabling individuals and smaller companies to bring novel ideas to market more quickly but are also optimizing workflows. Personally, these tools have streamlined my day-to-day work, saving me approximately 10 hours each week by automating routine tasks and enhancing productivity.

However, there are valid concerns about the impact of generative AI, particularly regarding its effect on the internet and the truth. As AI-generated content becomes more prevalent, there is a real risk of misinformation and the proliferation of fake information online. This not only threatens the integrity of the internet but also raises ethical questions that need urgent attention.

Interestingly, these challenges are creating new opportunities for AI ethics as a field. We’re likely to see significant job growth in areas focused on developing frameworks and tools to manage these risks, ensuring that AI is used responsibly and that the internet remains a trusted source of information.

While we are only getting started, the balance of benefits and challenges will ultimately shape the economic and social impact of generative AI. It’s an exciting time, but also one that demands careful consideration of the ethical implications.

How can companies effectively navigate the ethical considerations that come with the rapid advancements in AI technology? 

As an ex-McKinsey person myself, I feel compelled to steal some good advice and guidelines they have provided for this one:

Establish clear ethical guidelines: Companies should start by defining ethical principles that align with their values and business goals. These should cover critical areas such as bias and fairness, explainability, transparency, human oversight, data privacy, and security. For instance, ensuring that AI models do not inadvertently discriminate based on race, gender, or other protected characteristics is essential.Implement human oversight and accountability: It’s important to have a “human in the loop” to oversee AI decisions, particularly in high-stakes scenarios like financial services or healthcare. This ensures that there is always a human judgment applied to AI outputs, which can help mitigate risks associated with AI decision-making.Continuous monitoring and adaptation: Ethical AI isn’t a one-time effort. Companies should establish ongoing monitoring systems to track the performance and impact of AI models over time. This includes regular audits to check for biases or inaccuracies that might emerge as the AI system interacts with new data.Educate and empower employees: Building a culture that supports ethical AI requires educating employees across the organization about the importance of these issues. Providing training on ethical AI practices and ensuring that teams are equipped with the necessary tools to implement these principles is crucial for long-term success.

Generative AI is a whole new ballgame, and we still have a lot to learn, but these pillars provide a good start.

What are you excited about at Generative AI Summit Toronto, and why is it important to get together with other leaders like this?

I’m really excited about the opportunity to connect with a diverse group of AI professionals and thought leaders at the Generative AI Summit in Toronto. 

The event will feature cutting-edge discussions on the latest advancements in generative AI, and I’m particularly looking forward to the workshops and panels that provide opportunities to interact directly with experts. It’s important to gather with other leaders in the field to share insights, foster collaboration, and drive innovation in this rapidly evolving space.

Christie will be moderating at AI Accelerator Institute’s Generative AI Summit Toronto.

Join us on Novevember 20, 2024.

Get your tickets below.

Generative AI is no longer a futuristic concept—it’s a transformative force reshaping today’s most innovative industries. Companies like Klarna and J.P. Morgan are making bold moves by integrating Generative AI into their operations, challenging the status quo and enabling unprecedented efficiency and creativity.

This shift isn’t merely an incremental upgrade; it’s a paradigm change that allows organizations to automate complex processes, generate creative content, and make data-driven decisions more effectively than ever before.

Yet despite its clear potential, many large companies hesitate, caught in the throes of Clayton Christensen’s “Innovator’s Dilemma.”

They are torn between the safety of their profitable legacy systems and the uncertain but promising path of investing in disruptive technologies like Generative AI. For these companies, the risk extends beyond lagging behind competitors; it’s the danger of becoming irrelevant in a landscape that rewards agility and punishes complacency.

In today’s fast-paced market, comfort zones have become liabilities. Companies that cling to legacy approaches while ignoring the winds of change are playing a dangerous game—one that could end with them being outpaced, outperformed, and ultimately pushed out of the market.

Disruption is relentless: Comfort zones are a liability

Christensen’s “Innovator’s Dilemma” illustrates how companies often lose their edge by focusing on existing products while ignoring larger shifts. Generative AI represents one of these shifts, transforming industries with innovations that enhance efficiency and open new possibilities.

Klarna’s recent decision to move away from well-established SaaS platforms like Salesforce and Workday exemplifies this transformation. By developing an internal AI-driven solution, they are not only replicating but surpassing decades of customization and workflow automation offered by industry giants.

This bold move challenges the narrative of SaaS ‘stickiness’ and highlights how companies that remain in their comfort zones risk being outpaced by more agile competitors.

Consider how Blockbuster, a giant in its heyday, ignored the rise of digital streaming while Netflix evolved from a niche DVD rental service into a streaming powerhouse. Companies that fail to adopt Generative AI today risk a similar fate. Disruptive technologies don’t pause for established players—they redefine industries and leave behind those who can’t or won’t adapt.

Malcolm Gladwell’s “The Tipping Point” emphasizes that transformative shifts often begin subtly, with small, almost unnoticeable changes that eventually reach critical mass.

Many companies entrenched in their comfort zones overlook these initial signs, dismissing them as inconsequential until the tipping point is reached and transformation becomes unavoidable. Finding a balance between existing customer needs and innovation is essential for long-term survival.

Balancing current needs with future vision

While addressing current customer needs is important, it’s equally critical for companies to anticipate future market demands. Generative AI technologies often start small, catering to niche markets or solving problems not immediately apparent to mainstream customers.

However, as these technologies evolve, they can redefine entire industries. Gladwell discussed how niche innovations, initially overlooked or even ridiculed, can suddenly become the next big thing when they reach a tipping point, rapidly gaining acceptance and disrupting established markets.

Focusing solely on present needs can leave companies vulnerable when market dynamics shift. To stay competitive, leaders must balance immediate demands with a clear vision for the future, ensuring their strategies include investments in disruptive technologies like Generative AI.

Klarna’s pivot to Generative AI illustrates the importance of this balance. While traditional SaaS platforms had been integral to their operations, Klarna recognized the potential of AI to streamline processes and reduce complexity.

By standardizing workflows and leveraging AI, they’ve created a more agile solution that meets current demands while positioning themselves for future growth. This move underscores Gladwell’s point about niche innovations gaining rapid acceptance when they reach a tipping point.

AI implementation: Costs and complexity

Having a vision for Generative AI is only part of the equation; the real test lies in execution, where costs and complexities can become formidable barriers. Implementing Generative AI requires a comprehensive, multi-year strategy.

The expenses associated with AI software, infrastructure, and staff training are significant hurdles that can deter many organizations. According to a 2023 report by McKinsey & Company, companies investing in AI can expect to allocate 20-30% of their IT budgets toward AI initiatives.

Klarna’s success wasn’t just about adopting new technology; it involved reengineering their tech stack from the ground up and embracing standardization to reduce complexity.

This approach demanded a significant commitment but resulted in a more agile and cost-effective system. Their experience demonstrates that while the barriers to AI implementation are real, they can be overcome with a strategic, long-term vision.

Integrating Generative AI is not just about acquiring technology—it’s about embedding it into the organizational DNA and aligning it with strategic business goals. This involves substantial investments not only in technology but also in people and processes, requiring a commitment to long-term change rather than short-term fixes.

Organizations that hesitate because of these initial hurdles risk being left behind as others recognize the potential and reach the tipping point where Generative AI shifts from experimental to essential.

Established organizations often value stability and incremental improvements. Generative AI challenges these norms, requiring businesses to rethink how value is created and delivered. The psychological barrier—the fear of undermining one’s own success—can paralyze decision-making and lead companies to stick with what’s safe rather than explore new frontiers.

Success amid challenges: J.P. Morgan

Despite these challenges, companies like J.P. Morgan are successfully navigating this complex landscape. J.P. Morgan has launched an AI-powered chatbot for its research analysts, streamlining access to insights and data across the organization.

This initiative reflects a broader strategy to embed Generative AI within the company’s operations, enhancing decision-making and fostering a culture of agility and innovation. By taking a proactive approach, J.P. Morgan is not just adopting Generative AI—it’s transforming how it does business, setting a blueprint for other companies on how to integrate AI successfully.

While measuring success in Generative AI can be challenging due to the early nature of the technology, the initial benefits are already reshaping business operations. One significant hurdle is establishing clear ROI and KPIs, as many AI projects are still in exploratory stages.

However, a Deloitte survey found that over 50% of early AI adopters reported a positive return on their investment. Leaders need to invest with a long-term vision, understanding that while specific metrics may still be evolving, the transformative impact of AI is increasingly undeniable.

The flexibility of small players: How nimble newcomers are disrupting the status quo

While established companies are adapting, smaller, more agile newcomers are often best positioned to capitalize on Generative AI’s potential quickly. Without the burden of legacy systems and entrenched processes, these newcomers can experiment, adapt, and scale AI initiatives more effectively.

Companies like Writesonic and Gamma.app are leveraging Generative AI to reshape industries such as content creation and business communication. They exemplify how agile players can outmaneuver larger, slower competitors.

As Gladwell describes in “The Tipping Point,” these innovations can shift from fringe concepts to mainstream essentials, catching larger companies off guard when they reach that critical tipping point.

Klarna’s bold strategy doesn’t just signify a shift for one company; it poses critical questions for the entire SaaS industry. If AI enables enterprises to replace decades of deep integration with more agile, customized solutions, the traditional ‘stickiness’ of SaaS platforms is under threat.

This development forces CIOs and IT leaders to reconsider their reliance on established providers and explore the potential of in-house AI-driven solutions. The financial stakes are high, as enterprises could save millions annually by reducing dependence on costly SaaS products.

According to Gartner, organizations can reduce operational costs by 20 – 30% by 2025 through AI-driven efficiencies. Klarna’s example may well be the tipping point that accelerates a broader move away from traditional SaaS, emphasizing the urgent need for companies to adapt or risk obsolescence.

Ethical and social considerations

As companies embrace Generative AI, it’s crucial to address ethical and social considerations. Issues such as data privacy, security, and algorithmic bias can pose significant risks if not properly managed.

A 2022 survey by PwC revealed that over 55% of consumers are concerned about how companies use their personal data. Implementing robust data governance policies and ethical guidelines is essential to build trust with stakeholders and ensure compliance with regulations like GDPR.

Moreover, the impact of AI on the workforce cannot be ignored. While AI can automate routine tasks, it may also lead to job displacement. Companies should invest in retraining and upskilling employees to work alongside AI technologies, fostering a culture of continuous learning and adaptation.

Taking action: A roadmap for embracing generative AI

To move from theory to practice, companies must take deliberate steps to integrate Generative AI into their operations. Here’s how leaders can begin this transformative journey:

Gain executive buy-in: Executive-level support is critical for success.Conduct an AI readiness assessment: Evaluate your organization’s current capabilities, identify gaps, and set clear objectives for AI adoption.Develop a strategic AI roadmap: Align AI initiatives with business goals, prioritize use cases, and create a phased implementation plan.Start with pilot projects: Implement small-scale AI projects to demonstrate value, set measurable metrics, and iterate based on insights.Invest in talent and training: Upskill existing employees, hire specialized talent, and foster a culture of innovation.Address ethical and governance considerations: Establish ethical guidelines, implement governance frameworks, and engage stakeholders transparently.Leverage partnerships and collaborations: Collaborate with AI vendors, join industry consortia, and engage academic institutions.Monitor and measure impact: Set clear KPIs, conduct regular reviews, and scale successful projects.Plan for long-term sustainability: Stay informed on AI developments, budget for ongoing investment, and anticipate future needs.

By following this roadmap, companies can navigate the complexities of Generative AI adoption, mitigate risks, and position themselves for long-term success in an increasingly AI-driven world.

The path forward: Embrace generative AI or face extinction

The lessons from the “Innovator’s Dilemma” speak volumes: focusing solely on today’s successes without investing in disruptive technologies like Generative AI is a risky bet. AI isn’t just another tool; it’s a fundamental shift in how businesses operate.

Companies that fully integrate Generative AI into their operations will not only survive but thrive, setting the pace for their industries. In contrast, those who fail to adapt risk meeting the same fate as Blockbuster and BlackBerry—left behind in a world increasingly driven by AI that rewards the bold and punishes the complacent.

Jim Collins, in “Good to Great,” emphasizes that truly great companies continuously evolve and align their strategies with the future. Klarna’s decision to harness Generative AI reflects this principle, demonstrating proactive leadership and a commitment to innovation.

Their approach serves as a blueprint for other companies: not just to adopt new technology but to redefine their operations and strategies around it. Without such commitment, companies risk stagnation—going from good to gone.

Conclusion

For leaders, the message is simple: adapt, innovate, and lead, or risk becoming a cautionary tale. Klarna and J.P. Morgan’s transformations illustrate that the future belongs to those willing to embrace change and leverage disruptive technologies to their advantage. The decision isn’t just about adopting new technology; it’s about ensuring your company is poised to excel tomorrow.

As Generative AI continues to advance rapidly, the window of opportunity to lead is narrowing. By taking proactive steps—assessing readiness, developing strategic roadmaps, investing in talent, and more—companies can overcome barriers and seize the transformative potential of AI. Embrace the change because Generative AI won’t wait, and neither should you. The time to act is now. 

References

1. Christensen, C. M. (1997). The innovator’s dilemma: When new technologies cause great firms to fail. Harvard Business Review Press.

2. Treiber, M. (2023). Klarna’s bold move: What it means for the future of SaaS in the enterprise. IKANGAI. https://www.ikangai.com/klarnas-bold-move-what-it-means-for-the-future-of-saas-in-the-enterprise/

3. Gladwell, M. (2000). The tipping point: How little things can make a big difference. Little, Brown.

4. McKinsey & Company. (2023). The state of AI in 2023: Generative AI’s breakout year. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-in-2023-generative-AIs-breakout-year

5. J.P. Morgan. (2023). J.P. Morgan introduces AI-powered chatbot for research analysts. J.P. Morgan News. https://www.jpmorgan.com

6. Deloitte. (2022). State of AI in the enterprise, 5th edition. Deloitte. https://www.deloitte.com

7. PwC. (2022). Consumer intelligence series: Trusted tech. PwC. https://www.pwc.com

8. Collins, J. (2001). Good to great: Why some companies make the leap… and others don’t.

Want access to hundreds of hours of expert talks?

Sign up for our Pro+ membership and watch presentations from some of the world’s leading companies in AI.

That’s 100+ hours from all of our events in one convenient place.

As artificial intelligence (AI) continues to evolve, its deployment has expanded beyond cloud computing into edge devices, bringing transformative advantages to various industries.

AI inference at the edge computing refers to the process of running trained AI models directly on local hardware, such as smartphones, sensors, and IoT devices, rather than relying on remote cloud servers for data processing.

This rapid evolution of the technology landscape with the convergence of artificial intelligence (AI) and edge computing represents a transformative shift in how data is processed and utilized.

This shift is revolutionizing how real-time data is analyzed, offering unprecedented benefits in terms of speed, privacy, and efficiency. This synergy brings AI capabilities closer to the source of data generation, unlocking new potential for real-time decision-making, enhanced security, and efficiency.

This article delves into the benefits of AI inference in edge computing and explores various use cases across different industries.

Fig 1. Benefits of AI Inference in edge computing

Real-time processing

One of the most significant advantages of AI inference at the edge is the ability to process data in real-time. Traditional cloud computing often involves sending data to centralized servers for analysis, which can introduce latency due to the distance and network congestion.

Edge computing mitigates this by processing data locally on edge devices or near the data source. This low-latency processing is crucial for applications requiring immediate responses, such as autonomous vehicles, industrial automation, and healthcare monitoring.

Privacy and security

Transmitting sensitive data to cloud servers for processing poses potential security risks. Edge computing addresses this concern by keeping data close to its source, reducing the need for extensive data transmission over potentially vulnerable networks.

This localized processing enhances data privacy and security, making edge AI particularly valuable in sectors handling sensitive information, such as finance, healthcare, and defense.

Bandwidth efficiency

By processing data locally, edge computing significantly reduces the volume of data that needs to be transmitted to remote cloud servers. This reduction in data transmission requirements has several important implications; it results in reduced network congestion, as the local processing at the edge minimizes the burden on network infrastructure.

Secondly, the diminished need for extensive data transmission leads to lower bandwidth costs for organizations and end-users, as transmitting less data over the Internet or cellular networks can translate into substantial savings.

This benefit is particularly relevant in environments with limited or expensive connectivity, such as remote locations. In essence, edge computing optimizes the utilization of available bandwidth, enhancing the overall efficiency and performance of the system.

Scalability

AI systems at edge can be scaled efficiently by deploying additional edge devices as needed, without overburdening central infrastructure. This decentralized approach also enhances system resilience. In the event of network disruptions or server outages, edge devices can continue to operate and make decisions independently, ensuring uninterrupted service.

Energy efficiency

Edge devices are often designed to be energy-efficient, making them suitable for environments where power consumption is a critical concern. By performing AI inference locally, these devices minimize the need for energy-intensive data transmission to distant servers, contributing to overall energy savings.

Hardware accelerator

AI accelerators, such as NPUs, GPUs, TPUs, and custom ASICs, play a critical role in enabling efficient AI inference at the edge. These specialized processors are designed to handle the intensive computational tasks required by AI models, delivering high performance while optimizing power consumption.

By integrating accelerators into edge devices, it becomes possible to run complex deep learning models in real time with minimal latency, even on resource-constrained hardware. This is one of the best enablers of AI, allowing larger and more powerful models to be deployed at the edge. 

Offline operation

Offline operation through Edge AI in IoT is a critical asset, particularly in scenarios where constant internet connectivity is uncertain. In remote or inaccessible environments where network access is unreliable, Edge AI systems ensure uninterrupted functionality.

This resilience extends to mission-critical applications, enhancing response times and reducing latency, such as in autonomous vehicles or security systems. Edge AI devices can locally store and log data when connectivity is lost, safeguarding data integrity.

Furthermore, they serve as an integral part of redundancy and fail-safe strategies, providing continuity and decision-making capabilities, even when primary systems are compromised. This capability augments the adaptability and dependability of IoT applications across a wide spectrum of operational settings.

Customization and personalization

AI inference at the edge enables a high degree of customization and personalization by processing data locally, allowing systems to deploy customized models for individual user needs and specific environmental contexts in real-time. 

AI systems can quickly respond to changes in user behavior, preferences, or surroundings, offering highly tailored services. The ability to customize AI inference services at the edge without relying on continuous cloud communication ensures faster, more relevant responses, enhancing user satisfaction and overall system efficiency.

The traditional paradigm of centralized computation, wherein these models reside and operate exclusively within data centers, has its limitations, particularly in scenarios where real-time processing, low latency, privacy preservation, and network bandwidth conservation are critical.

This demand for AI models to process data in real time while ensuring privacy and efficiency has given rise to a paradigm shift for AI inference at the edge. AI researchers have developed various optimization techniques to improve the efficiency of AI models, enabling AI model deployment and efficient inference at the edge.

In the next section we will explore some of the use cases of AI inference using edge computing across various industries. 

Use cases

The rapid advancements in artificial intelligence (AI) have transformed numerous sectors, including healthcare, finance, and manufacturing. AI models, especially deep learning models, have proven highly effective in tasks such as image classification, natural language understanding, and reinforcement learning.

Performing data analysis directly on edge devices is becoming increasingly crucial in scenarios like augmented reality, video conferencing, streaming, gaming, Content Delivery Networks (CDNs), autonomous driving, the Industrial Internet of Things (IoT), intelligent power grids, remote surgery, and security-focused applications, where localized processing is essential.

In this section, we will discuss use cases across different fields for AI inference at the edge, as shown in Fig 2.

Fig 1. Applications of AI Inference at the Edge across different fields

Internet of Things (IoT)

The expansion of the Internet of Things (IoT) is significantly driven by the capabilities of smart sensors. These sensors act as the primary data collectors for IoT, producing large volumes of information.

However, centralizing this data for processing can result in delays and privacy issues. This is where edge AI inference becomes crucial. By integrating intelligence directly into the smart sensors, AI models facilitate immediate analysis and decision-making right at the source.

This localized processing reduces latency and the necessity to send large data quantities to central servers. As a result, smart sensors evolve from mere data collectors to real-time analysts, becoming essential in the progress of IoT.

Industrial applications

In industrial sectors, especially manufacturing, predictive maintenance plays a crucial role in identifying potential faults and anomalies in processes before they occur. Traditionally, heartbeat signals, which reflect the health of sensors and machinery, are collected and sent to centralized cloud systems for AI analysis to predict faults.

However, the current trend is shifting. By leveraging AI models for data processing at the edge, we can enhance the system’s performance and efficiency, delivering timely insights at a significantly reduced cost.

Mobile / Augmented reality (AR)

In the field of mobile and augmented reality, the processing requirements are significant due to the need to handle large volumes of data from various sources such as cameras, Lidar, and multiple video and audio inputs.

To deliver a seamless augmented reality experience, this data must be processed within a stringent latency range of about 15 to 20 milliseconds. AI models are effectively utilized through specialized processors and cutting-edge communication technologies.

The integration of edge AI with mobile and augmented reality results in a practical combination that enhances real-time analysis and operational autonomy at the edge. This integration not only reduces latency but also aids in energy efficiency, which is crucial for these rapidly evolving technologies.

Security systems

In security systems, the combination of video cameras with edge AI-powered video analytics is transforming threat detection. Traditionally, video data from multiple cameras is transmitted to cloud servers for AI analysis, which can introduce delays.

With AI processing at the edge, video analytics can be conducted directly within the cameras. This allows for immediate threat detection, and depending on the analysis’s urgency, the camera can quickly notify authorities, reducing the chance of threats going unnoticed. This move to AI-integrated security cameras improves response efficiency and strengthens security at crucial locations such as airports.

Robotic surgery

In critical medical situations, remote robotic surgery involves conducting surgical procedures with the guidance of a surgeon from a remote location. AI-driven models enhance these robotic systems, allowing them to perform precise surgical tasks while maintaining continuous communication and direction from a distant medical professional.

This capability is crucial in the healthcare sector, where real-time processing and responsiveness are essential for smooth operations under high-stress conditions. For such applications, it is vital to deploy AI inference at the edge to ensure safety, reliability, and fail-safe operation in critical scenarios.

Autonomous driving

Autonomous driving is a pinnacle of technological progress, with AI inference at edge taking a central role. AI accelerators in the car empower vehicles with onboard models for rapid real-time decision-making.

This immediate analysis enables autonomous vehicles to navigate complex scenarios with minimal latency, bolstering safety and operational efficiency. By integrating AI at the edge, self-driving cars adapt to dynamic environments, ensuring safer roads and reduced reliance on external networks.

This fusion represents a transformative shift, where vehicles become intelligent entities capable of swift, localized decision-making, ushering in a new era of transportation innovation.

Conclusion

The integration of AI inference in edge computing is revolutionizing various industries by facilitating real-time decision-making, enhancing security, and optimizing bandwidth usage, scalability, and energy efficiency.

As AI technology progresses, its applications will broaden, fostering innovation and increasing efficiency across diverse sectors. The advantages of edge AI are evident in fields such as the Internet of Things (IoT), healthcare, autonomous vehicles, and mobile/augmented reality devices.

These technologies benefit from the localized processing that edge AI enables, promising a future where intelligent, on-the-spot analytics become the standard. Despite the promising advancements, there are ongoing challenges related to the accuracy and performance of AI models deployed at the edge.

Ensuring that these systems operate reliably and effectively remains a critical area of research and development. The widespread adoption of edge AI across different fields highlights the urgent need to address these challenges, making robust and efficient edge AI deployment a new norm.

As research continues and technology evolves, the potential for edge AI to drive significant improvements in various domains will only grow, shaping the future of intelligent, decentralized computing.

Want to know more about how generative companies are using AI?

Get your copy of our Gen AI report below!

Artificial intelligence: The impact of hype, economics and law

Artificial Intelligence (AI) continues to be a subject dominated by hype across the globe. According to McKinsey’s technology trends outlook 2024, 2023 saw $36 billion of equity investment in Generative Artificial Intelligence, whereas $86 billion was invested in applied AI [1].

Currently, the UK AI market is worth in excess of £16.8 billion, with forecasted growth of over £801.6 billion by 2035 [2], reflecting the sizeable economic and technological traction AI is taking across sectors. 

Through the application of Computer Vision technology, for example, Marks and Spencer saw over 10 weeks an 80% reduction in warehouse accidents: just one of many ways in which AI is making a difference [3]. It however remains to be seen how effective coordinated governance will allow for innovation to thrive whilst maintaining cross-sector compliance.

Whilst the United Kingdom’s wider ambition is to be an AI Superpower, there has been continued debate and scrutiny about what constitutes effective AI regulation and how any continued iteration of such regulation would remain in alignment with key principles of law.

The United Kingdom’s vision for AI

The now-opposition government back in 2023 published its white paper, AI Regulation: A Pro-Innovation Approach. The plans outlined a principles-based approach to governance which was delegated to individual regulators.

While at the time it was thought that the UK’s approach and existing success in AI was down to effective regulator-led enforcement combined with technology-neutral legislation and regulations, the pace of AI highlighted gaps – both in opportunities and challenges – that would require addressing.

In the run-up to the 2024 UK General Election, regulation was of high importance in the Labour party’s manifesto under the “Kickstart economic growth” section, with the now-incumbent government seeking to strengthen AI regulation in specific areas. 

Keir Starmer – both prior to and post-election – emphasised the need for tougher approaches to AI regulation through, for example, the creation of a Regulatory Innovation Office (RIO) [4]. The aim of a Regulatory Innovation Office would, inter alia, set targets for technology regulators and monitor decision-making speed against core international benchmarks while providing guidance according to Labour’s higher-level industrial strategy. 

It, however, is not a new AI regulator and instead it will still be up to existing regulators to address AI within their specific fields. It is yet to be seen how a Regulatory Innovation Office would differ from the AI Safety Institute, the first state-backed organisation advancing AI safety established by the Conservative Government at the beginning of 2024 [5].

In addition to a new regulatory office, the planned creation of a National Data Library initiative aims to bring together existing research programmes and data-driven public services with strong safeguards and public benefit at its heart [4].

Wider issues in regulating AI

Government plans and economic potential aside, there are increasing expectations AI will solve the most pressing issues facing humanity. However, as a result of the pace there is a wider endemic issue of digital technologies challenging the functioning of law. In the long run, both a proportionate and future proof regulatory approach will be required regardless of where in the world approaches are developed.

To start with, defining AI is not straightforward: there is not a widely accepted definition, and considering various strands of sciences are affected either directly or indirectly by AI there is a risk of creating individualised definitions based on the specific field. Moreover, different types of intelligence could result in varying definitions of AI, even if the technological scope is not considered. 

Adding into the mixture the fields of Computer Science and Informatics – both not being directly mentioned in the AI Act, for example – demonstrates a lack of a commonly agreed technical definition of what AI is or could be. What follows from this are both general and theoretical questions and how this could be moulded into a legal definition. 

If, for example, both the principles of legal certainty and the protection of legitimate interests are taken, the existing definition of AI does not satisfy key requirements for legal definitions. The result instead is definitions that are ambiguous and debatable in practicability, creating a bottleneck in formulating domestic or even international AI regulation.

What is ultimately important is that any regulatory goal is aligned with the values of fundamental rights and the concrete protection of legal rights. Take the precautionary principle – an approach to risk management – which outlines that if a policy or action causes harm to the public and there is not a scientific agreement on the issue, that policy or action in question should not be carried out.

Applying this to AI becomes problematic as the effects in many cases are either not assessable just now or, in some cases, not at all. If then a risk assessment is carried out according to the proportionality principle – where the legality of an action is determined by the balance between the objective, means, and methods as well as the consequences of the action – where limited factual knowledge is obtainable, the actionability of such assessment becomes increasingly challenging.

Instead, it is the intersection of the technical functionality and the context of the application where a risk profile of an AI system can be obtained, but even then from a regulatory perspective these systems can vastly differ in risk profile.

Conclusion

The versatility of AI systems will present a range of opportunities and challenges depending on who uses them, what purposes they are used for and the resulting risk profiles. Attempting to regulate AI – which frankly speaking is an entire phenomenon with increasingly infinite branches of use cases – through a generalised Artificial Intelligence Act will not work.

Instead, deep-diving into the characteristics and the use cases of the differing algorithms and AI applications is more important and is strategically more likely to result in effective, iterative policymaking that is beneficial to society and innovation. 

Bibliography 

[1] McKinsey Tech Outlook 2024: www.mckinsey.com. (n.d.). McKinsey Technology Trends Outlook 2022 | McKinsey. [online] Available at: https://www.mckinsey.com/capabilities/mckinsey-digital/ our-insights/the-top-trends-in-tech#/. 

[2] AI Growth and Adoption: Hooson, M. (2024). UK Artificial Intelligence (AI) Statistics And Trends In 2024. [online] Forbes Advisor UK. Available at: https://www.forbes.com/uk/advisor/ business/software/uk-artificial-intelligence-ai-statistics-2024/. 

[3] M&S Computer Vision Example: Protex.ai. (2023). Marks and Spencer reduced incidents by 80% in their first 10 weeks of deployment. [online] Available at: https://www.protex.ai/case-studies/ marks-and-spencer#:~:text=This%20momentum%20led%20to%20an [Accessed 5 Sep. 2024]. 

[4] Labour Party Manifesto: The Labour Party. (2024). Kickstart economic growth – The Labour Party. [online] Available at: https://labour.org.uk/change/kickstart-economic-growth/#innovation [Accessed 30 Aug. 2024]. 

[5] AI Safety Institute: Aisi.gov.uk. (2024). The AI Safety Institute (AISI). [online] Available at: https://www.aisi.gov.uk [Accessed 30 Aug. 2024]. 

Interested in more from Ana? Make sure to give the articles below a read:

 The dawn of artificial intelligence (AI) was marred by a disturbing reality: systems designed for facial recognition consistently misidentified women and individuals with darker skin tones.

The repercussions extended beyond mere inconvenience; they were profoundly damaging, leading to wrongful arrests and the perpetuation of harmful stereotypes. This wasn’t a simple technical glitch. It was a glaring reflection of the predominantly male teams that built the technology, highlighting a fundamental flaw in the industry’s composition.

This narrative isn’t isolated. Across the tech landscape, a recurring pattern emerges: a lack of diversity that yields outcomes that are, at best, biased and, at worst, deeply harmful.

Despite its claims to innovation, the industry remains entrenched in an antiquated “bro culture” that marginalizes women and stifles diversity. The consequences of this exclusion reverberate far beyond the workplace, impacting the very technology that shapes our world.

The unseen costs of bro culture

The tech industry has long been dominated by a “bro culture” that elevates male perspectives and diminishes the contributions of women. This culture manifests in subtle and overt ways, from being interrupted or talked over in meetings to being passed over for promotions. The result is an industry where women are chronically underrepresented, especially in leadership roles.

However, the ramifications of this culture extend beyond the individual women affected. By sidelining women, the tech industry forfeits the innovation that springs from diverse perspectives.

Extensive research consistently demonstrates that diverse teams are more creative, more effective, and more likely to generate groundbreaking solutions. Yet, the industry remains stubbornly homogenous, clinging to a culture that is increasingly misaligned with its aspirations for progress.

A personal lens

Neja, a talented software engineer, shared her experiences navigating the challenges of a male-dominated tech environment. She recounted instances where she was the sole woman in team meetings, her ideas often dismissed or appropriated, while her male colleagues received recognition for her work. Neja’s story, unfortunately, resonates with countless women in the field.

To bridge the gender gap in tech and AI, we need a multifaceted approach that transcends good intentions. Concrete actions and accountability measures are essential to create an environment where women can flourish. In Neja’s words, “It’s not enough to open doors; we must build pathways that lead to the boardroom.”

Leadership accountability is paramount. Setting measurable diversity goals and regularly assessing progress are critical steps in shifting the culture and empowering more women to pursue careers in technology.

The imperative of diverse voices in AI development

The urgency for diversity is most pronounced in the realm of artificial intelligence. The World Economic Forum’s Global Gender Gap Report 2023 reveals a stark reality: only 22% of AI workers are women. This statistic underscores the profound gender disparity in the field and emphasizes the critical need to increase women’s participation.

AI systems are trained on massive datasets. If these datasets are biased, the AI will replicate and even amplify these biases. We’ve witnessed the damage this can inflict, from facial recognition software that misidentifies people of color to hiring algorithms that discriminate against women. These problems don’t originate from malice; they arise from the absence of diverse voices during the development process.

When women and other underrepresented groups are excluded from AI development, their perspectives and experiences are omitted from the data and algorithms.

This can lead to technology that fails to serve everyone equitably or, worse, actively harms marginalized groups. To build AI systems that are fair, equitable, and effective, it’s imperative to include diverse voices at every stage of development. It’s not just about mitigating bias; it’s about creating technology that works for everyone.

Women in leadership: Charting the course for technology’s future

Diversity in tech isn’t solely about numbers; it’s about influence. It’s insufficient to simply have more women in the room—they need to occupy leadership positions where they can shape the trajectory of technological advancements. Women leaders bring unique perspectives that are indispensable for ensuring that technology is developed with ethics, inclusivity, and societal impact in mind.

Without diverse women in leadership roles, the tech industry risks perpetuating a path where innovation benefits the few at the expense of the many. When women lead, they introduce fresh ideas, challenge assumptions, and champion practices that are more equitable.

This is particularly crucial in AI, where the stakes are high, and the potential for both positive and negative impacts is immense. Women leaders can guide the industry toward a future where technology is not only innovative but also ethical and inclusive.

Forging a more inclusive future

Addressing the gender imbalance in tech necessitates more than just well-meaning intentions. It demands concrete actions that foster an environment where women can thrive.

This includes implementing policies that promote diversity and inclusion, establishing mentorship and sponsorship programs, and holding leadership accountable for cultivating a supportive culture. It also entails elevating women into leadership roles where they can directly influence the future of technology.

Companies must re-evaluate how they promote and support women, ensuring they have access to high-visibility projects and clear pathways to leadership. It’s not enough to open doors; we must construct pathways that lead to the boardroom. Leadership accountability is crucial.

Setting measurable goals for diversity, regularly assessing progress, and celebrating the contributions of women in tech are key steps in transforming the culture and inspiring more women to pursue careers in technology.

A clarion call

The tech industry stands at a critical juncture. It can either cling to outdated norms and impede its own growth or embrace diversity and inclusion as the catalysts for innovation and success. Dismantling the barriers of bro culture isn’t just about achieving equality; it’s about creating superior technology that benefits all of humanity.

By elevating diverse women into leadership roles, we ensure that technology evolves in ways that are groundbreaking, ethical, and inclusive. The stakes are high—not just for women but for the future of the entire industry and society as a whole. This isn’t simply a matter of doing what’s right; it’s a strategic imperative for building a more just and equitable future.

Learn more about about bias in AI – check out the article below.