Google DeepMind CEO Demis Hassabis and Anthropic CEO Dario Amodei offered sharply framed but overlapping views on the pace, promise and risks of artificial general intelligence during a wide-ranging interview at this week’s World Economic Forum in Davos, Switzerland.
Amodei said progress toward systems that rival or exceed human capabilities across most domains could arrive faster than many expect, driven by AI’s growing ability to write code and assist with AI research itself. He described an emerging feedback loop in which models help build their successors, accelerating development, though hardware constraints such as chip manufacturing could slow that process.
Hassabis struck a more cautious note, arguing that while AI has advanced rapidly in areas such as coding and mathematics, breakthroughs in scientific creativity and hypothesis generation remain harder. Some domains, particularly in natural sciences, require experimental validation that limits how quickly AI systems can improve themselves.
Both CEOs emphasized that AI’s upside remains enormous, from accelerating scientific discovery to tackling disease and climate challenges. At the same time, they warned that risks are rising as systems become more autonomous. Amodei pointed to concerns ranging from labor displacement to misuse by bad actors and authoritarian states, while Hassabis stressed the need for global coordination on safety standards.
Despite differing timelines, both executives rejected fatalistic “doomer” narratives, arguing that careful governance, technical safeguards and collaboration could steer AI toward broad societal benefit rather than harm.
The discussion underscored a shared belief that the next few years will be decisive, not just for technology, but for how institutions adapt to it.