Shock and regret: scientists react to AI roiling life’s work
OpenAI and Anthropic each said last week that one of their AI models had disproved the exponential time hypothesis
Three computer scientists whose careers were built on an idea about the limits of computers have been shocked and saddened to learn that artificial intelligence may have disproven it. OpenAI and Anthropic each announced last week that one of their AI models had helped disprove the exponential time hypothesis (ETH), a claim that underpins hundreds of studies in computer science and biology.
The hypothesis, proposed by Russell Impagliazzo and Ramamohan Paturi of the University of California San Diego around 25 years ago, asserted that some problems are so difficult that computers could only solve them by trying nearly every possible answer. Ryan Williams of MIT, a leading researcher in the field, had long publicly expressed doubts about the hypothesis, and he now said he felt vindicated.
Russell Impagliazzo expressed disappointment in the discovery, stating that he would have preferred a human to discover it and explain the thought process behind the proof. He also expressed concern about the future of his profession, suggesting that much of the research discovery process may now be performed by AI, leaving humans to focus on more routine tasks.
Paturi, however, remained optimistic, believing that the frontier of research will continue to advance through human creativity and ingenuity. He urged young students to explore these new tools while developing their own judgment. Williams also expressed optimism about the future of the field, acknowledging that some researchers may choose to leave it due to the rapid pace of change, but emphasizing that the field remains worth exploring.
The scientists expressed concern over the concentration of powerful AI tools in the hands of companies, which decide when and how to release their findings. They stressed that publicly releasing a proof does not provide equal access to the tools that produced it, and that the community's assessment is still necessary. Williams emphasized that access to the AI model itself is the closest we can get to understanding the thought process behind the proof, which is often more valuable than the solution itself.
Written by urgent.news from The Hindu - Sci-Tech's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
