The world's top AI researchers squared off at the 2026 World Economic Forum in Davos this week, revealing deep divisions over how soon artificial general intelligence will arrive and what it means for humanity.

Dario Amodei, CEO of Anthropic, delivered the most aggressive predictions. He told the Davos audience that AI models will replace the work of all software developers within one year and reach "Nobel-level" scientific research in multiple fields within two years. Most strikingly, he predicted 50% of white-collar jobs will disappear within five years.

Google DeepMind CEO Demis Hassabis pushed back, saying current AI systems are "nowhere near" human-level AGI. He estimated genuine AGI is still 5-10 years away and requires "one or two more breakthroughs" beyond current large language model technology. In a joint appearance with Amodei, Hassabis gave only 50% odds that AGI might be achieved within the decade.

Yann LeCun, who recently departed Meta after 12 years, offered the sharpest critique. He called the AI industry "completely LLM-pilled" and argued that large language models will never achieve human-like intelligence. LLMs excel at language because "language is easy," LeCun explained, but they fail at physical-world reasoning—which is why domestic robots and level-five autonomous vehicles remain elusive.