As it turns out, OpenAI can really good on the computer.
On Monday, the company said it punctured the Exponential Time Hypothesis, or ETH, a nearly 30-year-old belief in the computer science community that some logic problems take a lot of time to grow. Hours later, Anthropic released its own research refuting ETH.
OpenAI said it had asked a group of outside computer scientists to organize an open review of the claims. The researchers have highlighted that their involvement is not an endorsement.
If the AI lab announcement goes ahead, it will likely reopen the hunt for better algorithms — something computer scientists think can’t get better.
The ultra-nerdy claims have become a topic of conversation on social media and in tech circles. Here’s how some thought leaders reacted to the OpenAI announcement:
Ryan Williams, professor of electrical engineering and computer science at MIT
Williams, one of three independent algorithm design experts who reviewed OpenAI’s claims, has not certified the company’s results.
He wrote in X that, if OpenAI’s claim turns out to be true, the new findings “should make every computer scientist give at least some weight to the possibility of P = NP,” bringing the company closer to another potential $1 million Millennium Prize Problem.
In more practical terms, Williams says the breakthrough he claims could lead to “faster algorithms.”
He previously said that he does not believe in ETH.
Gary Marcus, professor emeritus at NYU
Marcus, a frequent AI skeptic, wrote that this could be a “big deal if true,” in X.
John D. Cook, Ph.D. and technology consultant in applied mathematics
In a blog post, Cook said that OpenAI’s claims — while impressive — don’t ultimately answer the biggest open questions in computer science.
“The status of P = NP is still unknown,” he wrote. “While 3-SAT doesn’t take exponential time, it still takes more than polynomial time as anyone knows.”
Joshua Fonseca Rivera, AI Security Research Fellow at Anthropic
Rivera writes in X that OpenAI’s new claims are a reminder that humans can underestimate the intelligence of computers.
“The bitter lesson should be to make this not a surprise,” he wrote. “I know this logically, but…wow, we really don’t know anything.”
Timothy Gowers, full professor at the Collège de France and Fields Medal-winning mathematician
Gowers wrote of X about the development: “It is surprising and consequential that it is false.”
Dan Shipper, CEO at Saben
Shipper, the AI-pilled chief executive of a tech startup, writes in X what others are thinking…
“I regret not getting a Ph.D. in frontier mathematics,” he said. “(Not that I’ve thought about getting one but still).”