Dario Amodei, Demis Hassabis and Yann LeCun are shaping a high-stakes split over when and how human-level artificial intelligence might arrive, with direct implications for investment, national strategy and workforce planning. At the World Economic Forum in Davos in January 2026, Anthropic CEO Dario Amodei and Google DeepMind CEO Demis Hassabis publicly set out diverging forecasts for artificial general intelligence, while Yann LeCun, now leading his own startup AMI Labs after leaving Meta in November 2025, argued that the large language model paradigm itself cannot reach human-level intelligence. Their disagreement is less about a calendar date and more about fundamentally different conceptions of what intelligence requires, from coding and mathematics to grounded understanding of the physical world.
Amodei has taken the most aggressive near-term stance, telling the Davos audience that artificial intelligence models would replace the work of all software developers within a year and would reach Nobel-level scientific research capability in multiple fields within two years, and that within five years, fifty percent of white-collar jobs would disappear. He based these claims on what he observes inside Anthropic, describing engineers who have largely stopped writing code themselves and estimating that within six to twelve months, models will perform most tasks a software engineer does end to end. Demis Hassabis adopted a more tempered but still bold view, assigning a fifty percent probability that a system capable of exhibiting all the cognitive capabilities that humans possess will exist by the end of the decade, which puts his median expectation around 2029 or 2030. He stressed that today’s artificial intelligence is “nowhere near” human-level artificial general intelligence and argued that achieving it will require “one or two more breakthroughs,” especially in reasoning about the physical world and designing experiments rather than just solving verifiable problems in code and mathematics.
Yann LeCun, a Turing Award winner and former chief artificial intelligence scientist at Meta, has taken the most contrarian position by dismissing large language models as a dead end for reaching human-level intelligence and backing that view by founding AMI Labs, which is targeting a $3.5 billion pre-launch valuation and aiming to raise nearly $600 million. He argues that systems built only on predicting language tokens lack grounded world models, intuitive physics and causal understanding, and he promotes his Joint Embedding Predictive Architecture (JEPA) as an alternative path that learns abstract, multi-modal representations and predicts changes in state rather than the next word. The split among the three leaders highlights contrasting views of intelligence as either task performance, experimentally grounded scientific reasoning, or deeply embedded world models. For policymakers, including in Algeria, the divergence suggests avoiding bets on a single timeline, instead investing in artificial intelligence literacy and adaptable skills while watching for signs such as whether code generation advances from function-level assistance to full-system design in 2026, whether artificial intelligence begins designing experiments that yield new scientific knowledge, and whether AMI Labs can demonstrate superior physical reasoning compared with large language models.
