Regulating Artificial Intelligence is a 4D Challenge

The author is the founder of to siftAn FT-backed site about European start-ups

Leaders of G7 nations addressed a host of global concerns over sake-steamed Nomi oysters in Hiroshima last weekend: the war in Ukraine, economic resilience, clean energy and food security, among others. But he also added an additional item to his parting swag bag of good intentions: the promotion of inclusive and trustworthy artificial intelligence.

Recognizing the innovative potential of AI, the leaders expressed concern about risks to public safety and human rights. Launch of the Hiroshima AI Process, The G7 formed a working group to analyze the impact of generative AI models like ChatGPT and to head the leaders’ discussions later this year.

The initial challenges will be how to define AI, categorize its threats, and design an appropriate response. Is regulation best left to existing national agencies? Or is the technology so important that it calls for new international institutions? Do we need a modern equivalent of the International Atomic Energy Agency, established in 1957 to promote the peaceful development of nuclear technology and prevent its military use?

One can debate how effectively the institution of the United Nations has fulfilled that mission. Furthermore, nuclear technology involves radioactive material and large-scale infrastructure that are physically easy to contain. On the other hand, AI is comparatively cheap, invisible, pervasive and has infinite use cases. At a minimum, it presents a four-pronged challenge that must be addressed in more flexible ways.

The first dimension is discrimination. Machine learning systems are designed to discriminate, to spot outliers in patterns. This is good for detecting cancer cells in radiology scans. But it’s bad if black box systems trained on flawed data sets are used to hire and fire employees or authorize bank loans. Bias in, bias out, as they say. Restricting these systems to unacceptably high-risk areas, as the EU’s upcoming AI Act proposes, is a tough, precautionary approach. A more favorable approach may be to engage independent, specialist auditors.

Second, propaganda. As academic expert Gary Marcus warned the US Congress last week, generative AI could threaten democracy itself. Such models can generate plausible liars and fake humans at lightning speed and on an industrial scale.

The onus should be on technology companies themselves to watermark content and minimize disinformation, just as they suppressed email spam. Failure to do so will only increase calls for more drastic intervention. Precedent may be set in China, where a draft law puts responsibility for misuse of AI models on the creator rather than the user.

Third, clutter. No one can accurately predict what the overall economic impact of AI is going to be. But it seems pretty certain that it’s going to lead to the “commercialization” of swaths of white-collar jobs, as entrepreneur Vivienne Ming pointed out at the FT Weekend festivities in DC.

Computer programmers have widely adopted generative AI as a productivity-enhancing tool. In contrast, striking Hollywood screenwriters may be the first of many in a trade fearful of their core skills being automated. This messy story belies simple solutions. Nations have to adjust to social challenges in their own way.

Fourth, devastation. Incorporating AI into Lethal Autonomous Weapon Systems (LAWS), or killer robots, is a terrifying prospect. The principle that humans should always be in the decision-making loop can only be established and enforced through international treaties. This also applies to discussions about artificial general intelligence, the (possibly hypothetical) day when AI surpasses human intelligence in every field. Some evangelists dismiss this scenario as a distraction fantasy. But it certainly deserves attention from experts who warn of potential existential risks and call for international research collaboration.

Others may argue that trying to regulate AI is as futile as praying that the sun never sets. Laws always develop incrementally whereas AI is developing rapidly, But Marcus says he was encouraged by the bipartisan consensus for action in the US Congress. Perhaps fearing that EU regulators could set global standards for AI, as they did with data protection five years ago, US tech companies are also publicly supporting regulation.

G7 leaders should encourage a competition for good ideas. They now need to start a regulatory race to the top instead of presiding over a fearsome slide to the bottom.