Anthropic Co-Founder Tells IMF the Singularity Is Under Way

Jack Clark told international finance leaders in Bangkok that artificial intelligence has initiated an irreversible acceleration in scientific discovery and economic output.

Speaking at a high-level conference co-hosted by the International Monetary Fund and the Bank of Thailand, the Anthropic co-founder argued that the technological singularity is already underway. Clark used his address to urge central bankers, finance ministers, and multilateral economists to discard linear growth projections. Modern frontier models, he argued, are shifting from passive text engines into autonomous systems capable of conducting scientific research.

According to Clark, automated scientific progress represents the primary inflection point. Where human researchers previously spent months formulating hypotheses, synthesising literature, and screening candidates in disciplines such as chemistry, biology, and materials science, advanced models now compress these workflows into hours. The compounding feedback loop of machine learning systems improving subsequent generations of software and hardware is accelerating research cycles beyond historical benchmarks.

This scientific transition is anchored by unprecedented capital deployment. Frontier laboratories and hyperscale technology firms are allocating tens of billions of dollars to build computational infrastructure. Clusters are scaling from tens of thousands of specialised accelerators to facilities housing hundreds of thousands. These installations require gigawatt-scale power commitments, driving direct interventions in electrical grid planning and sovereign energy policy. Clark emphasised that the physical footprint of artificial intelligence is no longer theoretical, comparing the scale of infrastructure investment to historic national industrial mobilisations.

For macroeconomic planners, this rapid scaling introduces acute measurement challenges. Traditional econometric models assume a standard diffusion curve for general-purpose technologies, relying on established ratios between capital expenditure and labour. Artificial intelligence disrupts this framework by automating intellectual labour directly. Clark noted that cognitive tasks, including complex software engineering, regulatory drafting, and financial analysis, are increasingly executable by synthetic systems. As a result, the marginal cost of cognitive output is collapsing towards the raw cost of electricity and compute time.

Clark warned that multilateral institutions remain poorly equipped to track this transition. Central banks typically rely on lagging indicators to measure productivity and employment shifts. Generative models, however, deploy across global digital networks instantaneously once training and safety evaluations conclude. This creates an immediate gap between capability frontiers and regulatory comprehension, leaving policymakers to manage structural labour shifts with tools designed for twentieth-century manufacturing cycles.

Geopolitical divergence forms another core dimension of Clark’s thesis. The singularity, he noted, is arriving in an intensely concentrated manner. Advanced economies with domestic semiconductor supply chains and capital-rich hyperscalers capture early efficiency gains. Developing economies face asymmetric exposure. They risk the sudden devaluation of routine knowledge-work outsourcing, alongside prohibitive financial barriers to acquiring sovereign compute clusters. Nations without access to advanced silicon face dependency on foreign proprietary models for core economic and state functions.

This structural imbalance complicates international monetary governance. Fiscal authorities face the potential erosion of personal income tax revenues as intellectual tasks shift to corporate compute clusters. Meanwhile, central banks must balance inflation and interest rate policy amid violent swings in capital-to-labour ratios. Clark advised delegates to treat compute capacity not merely as commercial technology, but as critical economic infrastructure akin to telecommunications networks or national transport grids.

Clark framed the transition as an existential test of institutional agility. As automated discovery compounds, the lead time for governments to formulate responsive fiscal and labour policies continues to diminish. Can international financial institutions adapt their economic doctrines before automated discovery outpaces human regulatory capacity?