Artificial intelligence (AI) systems that rival or surpass human cognitive abilities could arrive within the next few years, Google DeepMind CEO Demis Hassabis and Anthropic CEO Dario Amodei said on Tuesday, warning that while job losses have yet to materialise at scale, labour markets may struggle to adapt once the technology accelerates.
Speaking at the World Economic Forum’s annual meeting 2026 in Davos, Hassabis said he still sees about a 50 per cent chance of AI systems that match all human cognitive capabilities by the end of the decade, while Amodei suggested the timeline could be shorter, driven by AI increasingly being used to build more powerful AI.
“I don’t think that’s going to turn out to be that far off,” Amodei said, referring to predictions that human-level AI could emerge by 2026-27. “We are now in terms of models that write code. I have engineers who say, ‘I don’t write any code anymore. I just let the model write the code.’”
Hassabis had a more cautious tone, saying progress has been “remarkable” in areas like coding and mathematics, but that key elements of scientific creativity remain elusive. “Coming up with the question in the first place, or coming up with the theory or the hypothesis. I think that’s much, much harder,” he said.
AI Building AI
A central uncertainty, both executives said, is whether AI systems can fully “close the loop” by autonomously designing, improving and deploying future generations of models, a development that could dramatically compress timelines.
“I think we might be six to 12 months away from when the model is doing most, maybe all, of what software engineers do end to end,” Amodei said, adding that once that happens, progress could move faster than many expect.
Hassabis said such self-improving systems are plausible in tightly defined domains, but face constraints in areas that require physical testing, hardware or real-world experimentation. “You’ve got hardware in the loop that may limit how fast the self-improvement systems can work,” he said.
Jobs: Lag Now, Disruption Later
Despite growing public concern, both executives said AI has yet to have a clear, measurable impact on employment, particularly outside software. Hassabis said recent labour market softness appears linked to post-pandemic over-hiring rather than AI displacement.
“In the near term, that is what will happen when a breakthrough technology arrives,” Hassabis said. “Some jobs will get disrupted, but new, even more valuable jobs will get created.”
However, he acknowledged early signs of pressure on entry-level roles. “I think we are going to see this year the beginnings of maybe impacting junior-level jobs and internships,” he said.
Amodei, who has previously warned that up to half of entry-level white-collar jobs could disappear within one to five years, said his view has not changed. “There’s a lag,” he said. “The labor market is adaptable, but as this exponential keeps compounding, somewhere between a year and five years it will overwhelm our ability to adapt.”
Inside Anthropic, he said, AI is already changing hiring needs. “On the more junior end and even the intermediate end, we actually need less and not more people,” Amodei said.
Governments Unprepared
Both leaders said governments are not yet grappling seriously enough with the economic consequences of rapid AI progress.
“I’m constantly surprised, even when I meet economists at places like this, that there’s not more work going on about what happens,” Hassabis said, pointing to the need to rethink how productivity gains and wealth are distributed.
Amodei said the pace of change adds urgency. “This is happening so fast and is such a crisis that we should be devoting almost all of our effort to thinking about how to get through this,” he said.
While Hassabis said he would prefer AI development to slow to allow societies more time to adjust, both executives acknowledged geopolitical competition, particularly between the United States and China, makes coordinated restraint difficult.
“I wish we had five to 10 years,” Amodei said. “But assume I’m right and it can be done in one to two years. It’s very hard to slow down when geopolitical adversaries are building the same technology.” |
