In a development that surprised many, an AI model recently disproved the Jacobian Conjecture, an 87-year-old open problem in mathematics that had stumped human experts. The feat came alongside AI solving a major question in quantum cryptography and making short work of Erdos problems. Terence Tao, widely regarded as the greatest living mathematician, turned to AI to help him understand the solution.
Yet for all these intellectual triumphs, the world looks much the same as it did before. Long-distance trucking employment is actually slightly higher than a decade ago, and while productivity has grown modestly, there has been no explosive economic transformation. This gap between capability and impact has prompted a growing debate among researchers and observers.
The Incremental Superintelligence
Ruxandra Teslo, a writer and analyst, recently noted the disconnect: “Walking around the world today one might notice that it is weirdly unchanged… To many, this is surprising.” She points to bottlenecks in governance and other frictions as the reason for the slow economic impact.
But a more radical hypothesis is gaining traction: that intelligence itself is subject to diminishing returns. Francois Chollet, an AI researcher known for measuring AI capabilities, argued in a series of tweets that intelligence is not an unbounded scalar stat like height. “Intelligence is a conversion ratio, with an optimality bound,” he wrote. “Increasing intelligence is not so much like ‘making the tower taller,’ it’s more like ‘making the ball rounder.’ At some point it’s already pretty damn spherical and any improvement is marginal.”
This perspective challenges the popular narrative of a coming singularity, where superintelligence bootstraps itself to godlike levels and transforms the universe. Instead, it suggests that even as AI gets better at specific tasks like math and coding, the overall impact on society may remain incremental.
Clifford Sosin, an investor and commentator, put it bluntly: “Superintelligence arrived. You probably didn’t notice, because it turned out to be kind of incremental… We were told to expect something bigger.” He notes that while AI can write excellent code and beat humans at a startling range of tasks, there has been no takeoff, no explosion.
Implications for Asia and the Indo-Pacific
For Asian economies that have bet heavily on AI-driven growth, this debate carries real stakes. In China, where the government has poured resources into AI development as part of its national strategy, the prospect of diminishing returns could reshape investment priorities. Similarly, in Japan and South Korea, where aging populations are looking to AI to offset labor shortages, the technology's incremental nature may temper expectations.
India, with its vast pool of IT talent, faces a different challenge: if AI's intelligence ceiling is real, the country's outsourcing industry may not be disrupted as quickly as feared, but the long-term competitive advantage of human intelligence could also erode more slowly. Meanwhile, Southeast Asian nations like Indonesia and Vietnam, which are positioning themselves as manufacturing hubs, may find that AI-driven automation arrives more gradually than anticipated.
The debate also touches on broader geopolitical dynamics. As the US and China compete for AI supremacy, the assumption that intelligence is a limitless resource has driven massive investments in compute infrastructure. If Chollet and others are right, those investments may yield diminishing returns, potentially altering the balance of power in the Indo-Pacific.
For now, the evidence is mixed. AI continues to improve at a rapid clip, but the world remains stubbornly unchanged. Whether this is a temporary bottleneck or a fundamental limit on intelligence itself is a question that will shape the region's future for decades to come.
