‘I’d rather shake a human hand’: Scientist’s bittersweet reaction as AI threatens to overturn his life’s work

SAN FRANCISCO, October 31 — A trio of computer scientists who built their careers around a hypothesis about the limits of what computers can achieve have learned that artificial intelligence may have disproved it. OpenAI and Anthropic, two leading AI companies, announced on Monday that one of their AI models had challenged the "exponential time hypothesis" (ETH).

This hypothesis posits that certain problems are so complex that computers can only solve them by attempting nearly every possible solution. The implications of this upheaval reverberated through academic circles, provoking reflection among researchers on the role of human expertise and the trajectory of their professions.

Russell Impagliazzo, Ramamohan Paturi, and Ryan Williams, the founders of the ETH, shared their reactions to the AI-driven challenge with AFP. The conversation has been streamlined for clarity and brevity.

Paturi expressed surprise at the development, noting that ETH has been a cornerstone of computer science research over the last three decades. Williams, meanwhile, confirmed he was a skeptic of the hypothesis and thus found the outcome reassuring. Impagliazzo, however, felt more pessimistic, wishing for a human to have discovered the result and for the satisfaction of shaking the proponent's hand.

Looking ahead, the trio discussed the ramifications for human researchers. While Impagliazzo harbored doubts about the pace of change, Williams and Paturi were optimistic about the potential for human ingenuity to adapt and thrive. They cautioned, however, about the potential for unequal access to AI tools and the importance of maintaining transparency in their development.

For ordinary people, the immediate impact of this result remains theoretical. Yet, should the findings hold, it could pave the way for more efficient computing methods across various fields. The researchers emphasized the importance of continued human involvement in the research process, particularly in the explanation and justification of AI-generated results.