OpenAI and Anthropic Challenge a Core Computer Science Assumption

OpenAI released solutions to three major math problems, including a surprising rejection of the Exponential Time Hypothesis (ETH), a key assumption in computer science. The ETH suggests that some problems require exponentially more time to solve as they grow in complexity. Both OpenAI and Anthropic claim their AI models have found algorithms that solve these problems faster than previously thought possible.

The Exponential Time Hypothesis and Its Importance

The Exponential Time Hypothesis (ETH) is a fundamental concept in computer science, closely related to the P versus NP problem. The P versus NP problem asks whether problems whose solutions can be quickly verified (NP) can also be quickly solved (P). A classic example is Sudoku: verifying a solution is easy, but finding one is time-consuming. The ETH supports the assumption that P ≠ NP, meaning not all verifiable problems can be quickly solved. It posits that some problems require exponential time to solve as they grow in complexity.

The ETH was first published in 1999 by UC San Diego computer scientists Russell Impagliazzo and Ramamohan Paturi. It has since become a cornerstone of computational complexity theory, influencing how researchers understand the limits of computation. The hypothesis has been widely accepted and incorporated into numerous scientific papers and practical applications.

The Claims by OpenAI and Anthropic

OpenAI and Anthropic have both claimed that their AI models have developed algorithms that solve problems in 'sub-exponential' time, meaning faster than the speeds previously thought possible. If true, these claims could refute the ETH and potentially open up a path for algorithms that generally perform P = NP. This would be a monumental shift in computer science, as it would imply that problems currently thought to be intractable could be solved efficiently.

The process of releasing these findings has been structured differently for each of the three disciplines involved. Mathematicians received the work on GitHub, computer scientists formed a working group and asked for the proofs to be shared publicly, and physicists formed a working group to determine how their result will be released. This collaborative approach aims to ensure that the scientific community can thoroughly review and understand the implications of these findings.

The Implications for Computer Science and Cryptography

The potential refutation of the ETH has profound implications for computer science. Many scientific papers over the past decade rely on the ETH, and if the AI labs' claims hold up, these papers will need to be revisited. The findings could also affect cryptographic systems, as faster algorithms might make certain encryption methods vulnerable. However, OpenAI has reportedly checked and is confident that their findings do not currently break any cryptographic systems.

MIT professor Ryan Williams, who is organizing the computer science community's response, has noted that the implications of these findings are significant. He suggested that every computer scientist should put a little more weight on the plausibility of P = NP, although the final answer is still uncertain. The findings call into question many of the intuitions researchers had about the tractability of hard NP problems.

The Broader Impact on Scientific Discovery

This new AI paradigm of scientific discovery is still a work in progress. While there are concerns, particularly about training the next generation of scientists, there is also optimism. AI is forcing researchers to be more ambitious and to think more deeply about the kind of scientific questions they want to ask. This could lead to a new era of scientific exploration and innovation.

For everyday people, the implications are significant. If the ETH is refuted, it could lead to breakthroughs in various fields that rely on computational complexity, from cybersecurity to artificial intelligence. It might also change how we approach problem-solving in general, potentially leading to more efficient algorithms for a wide range of applications.

What You Can Do Today

While this discovery is still being reviewed by the scientific community, you can stay informed by following updates from reputable sources like TechCrunch AI. If you're interested in the technical details, you can explore the open letter released by MIT professor Ryan Williams and other computer scientists, which calls for feedback and discussion on the norms around AI contributions to scientific work. You can also participate in the ongoing discussions and conferences organized to review these findings.