Anthropic CEO Dario Amodei said that artificial intelligence could surpass humans in most tasks in the coming years. In his opinion, this will create serious risks for labor markets and social institutions.
Speaking at a panel at the World Economic Forum in Davos, Amodei noted that the timeline for the emergence of artificial general intelligence (AGI) is shrinking. At the same time, the window for adapting public policy is quickly closing, the expert believes.
According to him, AI systems could reach or exceed human levels within the next five years. Amodei stressed that he finds it difficult to imagine a scenario in which this process would take decades.
He cited the use of AI to automate its development as one of the factors accelerating this process. At Anthropic, he said, engineers are increasingly moving from writing code to editing the results generated by models.
Google DeepMind CEO Demis Hassabis expressed a more cautious position. He noted that automation is easier in programming and mathematics than in the natural sciences, where results require experimental verification.
At the same time, Hassabis estimated the probability of AGI appearing by 2030 at approximately 50%. The expert pointed to the gap between the growth of computing power and the ability of AI to create fundamentally new theories.
Both leaders agreed that entry-level office workers will be the first to feel the impact. Amodey previously predicted that up to half of such positions could disappear within five years.
The participants in the discussion identified the key risk as the unpreparedness of states for rapid economic and social change. According to Amodey, in the absence of coordinated action, the speed of AI development could lead to systemic failures.
We previously reported that the head of Anthropic called the export of Nvidia chips to China “analogous to the supply of nuclear weapons.”


