Anthropic co-founder Jack Clark said on Oct. 17 that an AI-driven "singularity" has already started, telling hundreds of government and financial officials at the International Monetary Fund (IMF) and World Bank annual meetings in Bangkok that the technology could reshape economic output on a scale that now sounds extreme.
Clark cited simulations from Anthropic’s economics team and said that, under specific conditions, U.S. gross domestic product could reach $44 trillion by 2040. He also urged governments to cooperate on what he called "unusual and serious rules" for the technology and the companies developing it.
The remarks came during the Per Jacobsson Foundation lecture held alongside the IMF annual meetings. The talk, titled "Inside the Event Horizon," was moderated by former Reserve Bank of India Governor Raghuram Rajan, and the full video has been published on the IMF’s official YouTube channel.
Clark says scientific progress is set to speed up across the board
Clark said researchers and science fiction writers have long predicted that AI would sharply accelerate scientific and technological progress, to a point where humans would struggle to keep pace. "I believe this year we have gotten strong evidence that it has started," he said.
He referenced science fiction author Vernor Vinge’s 1993 description of the singularity and then said plainly that "the singularity has arrived." In Clark’s view, the central effect is straightforward: over the next few years, it will not only be AI that advances faster, but science as a whole.
He said that shift could be disorienting. It could also create enormous wealth and enormous danger. Clark described a future in which AI systems can be replicated at scale as long as enough computing power is available, allowing each human scientist to supervise hundreds or even thousands of "virtual scientists." Citing a UNESCO estimate of roughly 10 million scientists worldwide, he asked the audience what the world would look like if those 10 million people were directing 1 billion virtual scientists.
Anthropic model points to a $44 trillion U.S. economy in 2040
Clark said Anthropic’s economists modeled economic outcomes across different AI development paths. Under assumptions used inside AI labs, where models can autonomously perform a wide range of work and spread quickly through the economy in the next few years, "under these sort of singularity-like conditions, one of the models predicts a huge transformation in the economy, where U.S. GDP could reach $44 trillion by 2040."
Clark said he understood that the number sounded "very crazy," but argued that it follows from sustained technological improvement. If GDP were to grow that much, he said, economic theory suggests the expansion would need to be supported by new industries and new inventions, which is what AI-led scientific research could produce.
ABMedia noted that Reuters and CNBC described this section as implying that U.S. economic output could rise by nearly 50% before 2030. According to the IMF’s official video, however, Clark referred to 2040, not 2030.
In the discussion after the lecture, India’s Chief Economic Adviser V. Anantha Nageswaran separately cited an extreme scenario from Anthropic’s economic scenarios research released in September. In that case, the U.S. economy expands by one-third by 2030, labor’s share of income falls from about 60% to 45%, and knowledge workers’ wages end up more than 10% below the baseline.
Biology research should face the strictest oversight, Clark says
Clark said that using AI for science does not mean it should be allowed to operate without supervision. He singled out biology research as an area where "we need the strictest human oversight and transparency to maintain safety." He added that AI research itself raises the same issue, and that labs such as Anthropic need to decide how much oversight to impose on AI-led research.
He also said that after building data centers and power capacity, countries should next invest in scientific equipment and robotics. The countries that establish that supply chain first, he said, stand to benefit the most.
During the Q&A, Clark was asked whether humans should still hold the reins if AI becomes more capable than people. He referred to public comments by Anthropic CEO Dario Amodei, who has said it may be necessary to limit the intelligence ceiling of such systems. Clark said, "My view is that we are fully in control of this technology right now, but it is getting stronger much faster than our intuition suggests. We only have a few short years left to act."
Jobs could be hit hard in some sectors
On employment, Clark said occupations with high AI exposure could lose jobs, with customer service and call center work as examples. In some scientific fields, by contrast, AI could raise productivity and even increase employment.
He also said his models cannot predict what jobs new industries will create, but they do point to the possibility of very severe labor-market disruption.