By Dr Michael G. Kollo

Straddling the razor-thin line between artificial intelligence's potential and peril is akin to a high-stakes, global-scale circus act. On one side, AI is the prodigious juggler, flipping industries from health to finance to climate science to transportation into a higher orbit of efficiency and progress. It's a ticket to the big top for companies seeking competitive advantage.

However, every circus has its darker corners, and AI is no exception. As it snakes through society, AI may unleash a carousel of unintended consequences. The haunting melodies of fundamental rights breaches and social disruptions, echoed from AI's unchecked usage, can't be drowned out by the razzle-dazzle of progress.

Answering this discordant tune requires a carefully choreographed performance. The EU has been twirling towards a 'people-first' stance on AI regulation. They're busy stitching together a safety net of legislation that aims to make AI systems trustworthy for the public while coaxing businesses to continue fostering their growth.

But this high-wire act of tech advancement versus legislative control is no simple feat. Maintaining equilibrium between reaping AI's bountiful harvest and keeping its gnarlier tendrils in check is not just a regulatory act—it's a high-stakes societal juggling act. Holding steady on this tightrope demands a keen gaze and cat-like agility.

Stepping into this whirlwind of activity is the European Parliament and Council, recently making headlines with their expedited efforts in AI regulation. They've become a lighthouse in the fog for other governments, charting the murky waters of Generative AI and its potential impacts.

From the regulatory grandstand, AI systems, be they statistical, regular AI, or enigmatic Generative AI, are tagged as high-risk, earning them a spotlight of regulatory oversight. This classification hangs not on their mechanics (like supervised or unsupervised systems) but on their intended roles. Thus, the circus of AI regulation continues, a thrilling spectacle of technology, society, and the regulatory dance between them.

What is a high-risk AI system from a regulatory point of view?

In the European Union, some AI technologies get labelled as "high-risk." These systems could potentially harm people's health, safety, or basic rights. They're identified not just by what they do but also by their intended use and how they're used.

High-risk AI systems play roles in many important areas. They help manage critical infrastructure and are used in education, job training, employment, and credit scoring. They're also involved in immigration processes, law enforcement, and democratic activities like voting. Simply put, these systems can significantly affect our daily lives and society as a whole. So, it's clear why they're under special scrutiny by the EU.

How should high-risk AI systems act?

AI systems with high-risk tags must be architected and evolved to be tamed and controlled by human supervision. The purveyor of such a system carries the responsibility of determining suitable measures for human monitoring before the system's induction into the market or its activation. This facet of oversight becomes a substantial challenge for Generative AI—given that it largely relies on unsupervised models that churn out multifaceted nonlinear outcomes.

Moreover, high-risk AI systems ought to don the CE insignia, signalling their compliance with EU regulations. This marking smooths their unhindered movement within the boundaries of the European Union's single market—a crucial step in fostering technological advancement whilst keeping risks in check.

When would generative AI systems be considered high-risk AI systems?

A generative AI system gains the undesirable tag of "high-risk" when its application or manner of use holds the potential to significantly jeopardise health, safety, or basic human rights.

Such a situation usually arises when the AI system churns out content, prognoses, suggestions, or choices that substantially impact its interaction milieu—physical or digital. Simply put, a generative AI system operating in circumstances that could have a wide-reaching detrimental effect or disturb the routine flow of social or economic activities, is poised to earn the "high-risk" label.

Consider, for instance, an interview algorithm that leverages hefty language models to sort applicants for jobs or scholarships. This could potentially land in the "high-risk" category. Likewise, an AI that crafts credit risk parameters, partially banking on language models to classify written information, may be deemed a high-risk system—particularly if the credit score in question erects capital access barriers. In this regard, the AI realm presents a new battleground for risk and regulation.

The unstable path forward

In the unfettered realm of innovation, generative AI is at the cutting edge, pushing the envelope for regulators while turbocharging our pursuit to decode and steer the impacts of algorithms on our economy and society. The path to effective regulation is shrouded in ambiguity, marred by the disappointing attempts to regulate social media platforms and the ensuing societal fallout.

With a potential recession on the horizon, today's economic climate is raising anxiety levels around implementing novel AI systems. In such volatile times, small perceived risks can balloon into overarching fears - fear of change, fear of job security, and fear of dwindling human control. Buzzwords like 'ethical AI' and 'responsible AI' start to serve more as rallying cries against this fear rather than guiding principles for technological advancement.

Regulation, unfortunately, isn't the panacea for these fears. At best, it can pump the brakes on deploying these transformative technologies. As AI continues to be a strategic ace in the global geopolitical game, it's unlikely that these proposed regulations will touch its use in military or intelligence operations. Military agencies typically receive exemptions, with law enforcement agencies having certain conditional liberties.

This shift points to an evolving interpretation of privacy and ethics – it's not about a universal application of these principles but rather about their tailored application to protect a workforce and a domestic economy. The challenge, it seems, isn't just in regulating AI but in reframing the very values we hope to uphold in this brave new world.

Hear more from Dr Michael Kollo and other industry leaders at the Generative AI Summit and gain access to cutting-edge insights and strategies that will put your organisation ahead of the curve. Learn more.

To access the detailed conference program, download the brochure here.