What's wrong with calling it an 'AI race'? A lot-and India should reject this notion outright

None of this, however, means that artificial intelligence technology is a race. A few drivers could well race each other on the highway, but nothing stops other motorists from getting to their own destinations in their own time.

When confronted with evidence that the atomic and space ‘races’ of the previous century neither stopped new entrants nor conferred decisive advantages on anyone, proponents of a new race argue that this time is different because AI can build better AI on its own and this recursive self-improvement sets this technology apart.

This argument is strong, but must satisfy a few conditions to be persuasive. The compounding of capabilities must be disproportionate, faster than the pace at which it leaks to others and must escape the law of diminishing marginal returns. The hype from genuinely impressive achievements is currently standing in for real proof that we could have an AI to end all AIs.

Indeed, a race must fulfil all three conditions: one, there must be a definite finish line; two, crossing it first must entail a big reward; and three, this should not be available to other finishers. None of these is true in the case of AI. Some claim AGI is the finish line, but few can define what exactly it means. It is unclear if the first to reach that yet-to-be-defined mark will be the only one to reap its benefits, not least because most people will be satisfied with any one of the lesser models available to them.

We are likely to find that AGI, if and when we approach it, is more a moving horizon than a finish line. Those who think they are running a race might discover that they are actually on an accelerating treadmill.

What’s wrong with calling it a race? Won’t feverish competition drive innovation? Well, it would be one thing if it were only private investors putting their money into these ventures. What is happening is that governments around the world are allocating public funds, land, water and energy resources to participate in the race. All this might be better utilized to achieve economic growth, national power and prestige if clearer thinking prevailed.

The race metaphor gives the subject a strong security and geopolitical colour: generals, spooks and politicians gain undue influence on technical and economic decisions. We are already seeing this today in policies of yards, fences, export controls and sanctions, hitting not only AI, but the technology sector as a whole.

The security state displaces the regulatory state. Competition watchdogs and safety regulators are pushed aside. It is amusing to see regular headlines of AI labs postponing the release of their next model due to concerns over safety. The US does not allow aircraft manufacturers, pharmaceutical companies or even packaged food makers to put products in the market before safety certification. If AI labs get a pass, it is because the race metaphor allows them to get away with it.

India should reject the notion of an AI race. There is a case for publicly financed AI development, though not because India sees itself in a race with the US and China. Public policy should focus on rapid adoption and diffusion, which are the pathways of advantage for India.

Fellow Mint columnist Rahul Matthan calls it absorption. “We have been made to believe that those who build the most advanced AI models will reign over those who do not,” he writes in a new essay for Observer Research Foundation, yet “the value of a new technology is only ever realised when it is deployed into the world.”

This is the most important point as far as India is concerned. His “absorption frontier” is the point at which a sector has re-organized itself enough to turn abundant intelligence into productive gains.

It took decades for electricity to get there. AI will be faster, but public policy can accelerate the process. Countries that have built digital public infrastructure, Matthan argues, have already laid down the first layers of the frontier. India should build the infrastructure for absorption that will remain long after the bubble is gone.

In this month’s Foreign Affairs, Ruchir Sharma reminds us that “The economist William Nordhaus has shown that only two percent of the value created by innovations of the past, including railroads, airlines, and the Internet, went to the companies that built these networks. The rest went to the consumers and businesses that used the inventions.”

Tailpiece: “Aisa toh aadmi life mein doh hi time bhaagta hai... Olympic ka race ho ya police ka case ho” (A man runs like this only two times in his life… in an Olympic race or a police case). So said Anthony Gonsalves.

Nitin Pai

Nitin Pai is co-founder and director of the Takshashila Institution, an independent centre for research and education in public policy.He has been writing “The Intersection” column in Mint since February 2019, interpreting contemporary issues connected by geopolitics, technology, economics, science and philosophy.His current research includes economic statecraft, technology geopolitics and strategic studies. He teaches international relations, public policy and ethical reasoning at Takshashila’s graduate programmes.He is the author of "Nitopadesha: Moral Tales for Good Citizens" (Penguin Random House, 2023) and the co-editor of "India's Marathon: Reshaping the Post-Pandemic World Order", published in 2020.Pai spent over a decade in the Singapore government in the areas of broadband development and technology foresight. He has also worked with SingTel's international connectivity business and undersea cable projects.He was a gold medalist from the National University of Singapore’s Lee Kuan Yew School of Public Policy, an undergraduate scholar at Nanyang Technological University (NTU), and an alum of National College, Bangalore.