How AI will reshape the rules of business

Over the past few weeks, there have been many important developments in the global debate on AI risks and regulation. From both the US public hearings on OpenAI by Sam Altman and the announcement of revised AI legislation by the EU, the emerging theme is demand for more regulation.

But to the surprise of some, there is consensus among governments, researchers and AI developers about the need for this regulation. In testimony before Congress, OpenAI CEO Sam Altman proposed creating a new government agency to issue licenses for large-scale AI model development.

He offered several suggestions on how such bodies could regulate the industry, including “a combination of licensing and testing requirements,” and said companies like OpenAI should be independently audited. rice field.

But while there is growing agreement about risks, including potential impacts on people’s work and privacy, there is still uncertainty about what such regulation should look like and what potential audits should focus on. little agreement has been reached. The inaugural Generative AI Summit, held by the World Economic Forum, brought together AI leaders from business, government, and research institutions to drive collaboration on how to navigate these new ethical and regulatory considerations. bottom. Two important themes emerged there.

The Need for Responsible AI Audits

First, we need to update the requirements for companies developing and deploying AI models. This is especially important when asking what “responsible innovation” actually means. The UK is leading this debate, with the government recently providing guidance on AI through five core principles, including safety, transparency and fairness. And a recent study from the University of Oxford highlights that “LLMs such as ChatGPT urgently need to update their concept of responsibility.”

A central factor in this push for new responsibilities is the growing difficulty of understanding and auditing the new generation of AI models. To see this evolution, we can compare “traditional” AI and LLM AI, or Large Language Model AI, in the example of recommending job candidates.

When traditional AI is trained on data that identifies employees of a particular race or gender at higher-level jobs, it can be biased in recommending people of the same race or gender for jobs. There is a nature. Fortunately, this can potentially be detected or audited by inspecting the data used to train these AI models and the recommendations output.

With new LLM-powered AI, this kind of bias audit is becoming increasingly difficult, if not impossible, to test for bias and quality. Not only do we not know what data a “closed” LLM was trained on, but conversational recommendations can introduce bias and more subjective “hallucinations”.

For example, if you ask ChatGPT to summarize a presidential candidate’s speech, who decides if it’s a biased summary?

Products that include AI recommendations therefore take into account new responsibilities, such as the traceability of the recommendations, such as ensuring that the models used in the recommendations can actually be audited for bias rather than just using LLM. is more important than ever.

Key to the new AI regulations in HR are the boundaries of what counts as recommendations or decisions. For example, New York City’s new AEDT law promotes biased audits of technology, especially those related to hiring decisions, such as those that can automatically decide who to hire.

But the regulatory landscape is rapidly evolving beyond how AI makes decisions to how it is built and used.

Transparency in communicating AI standards to consumers

A second important theme emerges here. It is the need for governments to define clearer and broader standards for how AI technology should be built and how it should be made clear to consumers and employees.

At a recent OpenAI hearing, IBM Chief Privacy and Trust Officer Christina Montgomery emphasized the need for standards to ensure consumers are aware of their chatbots every time they use them. . This kind of transparency about how AI is developed and the risks of exploiting open source models is key to her recent EU AI bill consideration of LLM APIs and the ban on open source models.

The question of how to control the spread of new models and technologies will require further discussion before the trade-off between risks and rewards becomes clearer. But as the impact of AI accelerates, it is becoming increasingly clear that standards and regulations are becoming more urgent, and so is awareness of both risks and opportunities.

Impact of AI Regulation on HR Teams and Business Leaders

Perhaps the earliest to feel the impact of AI is the HR team. HR teams are being asked to both manage the new pressures to provide employees with upskilling opportunities and to provide management with coordinated forecasts and workforce plans for new skills needed. increase. To adapt business strategy.

At the two recent WEF Summits on Generative AI and the Future of Work, I spoke with AI and HR leaders, as well as policy makers and academics, about the need for all businesses to drive responsible AI adoption and awareness. We talked about a new consensus. The WEF has released its Future of Jobs report, which highlights that 23% of jobs will change over the next five years, creating 69 million jobs and excluding 83 million. This means that at least 14 million jobs will be deemed at risk.

The report also found that by 2027, 6 in 10 workers will not only need to change their skill sets to get a job, but will also need to be upskilled and reskilled. , also stressed that only half of its workforce appears to have access to adequate training opportunities today.

So how should teams continue to engage their employees in AI-accelerated transformation? By carefully considering how to create a compliant, connected set of people and technology experiences that improve.

A new wave of regulation is shedding new light on how we consider biases in relationship decision-making, such as talent, as these technologies are being adopted by people both inside and outside the workplace. , its responsibility is greater than ever. Empower business and HR leaders to understand both the technology and regulatory landscape so they can commit to driving responsible AI strategies for their teams and businesses.

Sultan Saidov is president and co-founder of Beamery.