RESOLVING THE LEDGER
This is one outlet's own report from Scientific American — the article as it was filed. Other outlets are covering the same event; open the full story to compare every source side by side.
See the full story · 1 sources
Two research efforts gave AI the same open problem. Their papers were submitted three hours apart
An M.I.T. Ph.D. student and two University of California system cryptographers used GPT-5.6 Sol Ultra in different ways, raising new questions about independent discovery and scientific credit
Last Thursday Massachusetts Institute of Technology graduate student Seyoon Ragavan sent a new quantum cryptography proof to his friend Yao-Ting Lin, a doctoral student at the University of California, Santa Barbara. Lin could not look at it right away, though. He was in a meeting with his adviser, U.C.S.B. professor Prabhanjan Ananth , where he heard about another proof of the same result. After the meeting, he e-mailed back to Ragavan, "We are definitely living in strange times."
That day the researchers submitted two preprint papers to arXiv.org. One was by Ragavan. The other was by Ananth and University of California, Los Angeles, professor Amit Sahai . Both papers credited OpenAI's newly released GPT-5.6 Sol Ultra with finding the core ideas behind their proofs and construction.
The convergence was striking, though not entirely spontaneous. Ragavan and Sahai had heard the same open question posed earlier that month at the Simons Institute for the Theory of Computing at the University of California, Berkeley. They pursued it separately, using the same model through different workflows. (Disclosure: The talk at the Simons Institute was given by a researcher who was then a member of the same laboratory as the author of this story.)
If you're enjoying this article, consider supporting our award-winning journalism by subscribing . By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.
Neither paper has been peer-reviewed, and the result concerns a specialized corner of quantum cryptography known as unclonable encryption. But the near collision offers an unusually vivid glimpse of how AI is changing scientific research. When separate researchers can direct the same model toward the same question, results may emerge almost simultaneously—raising new questions about what counts as independent discovery and who deserves credit.
"Now the general mentality is: if someone mentions an open problem, the first thing is to see if GPT solves it," Ananth says. "A lot of problems that we didn't know how to solve are going to get solved in the very near future."
AI has spent months picking off old math problems—an 80-year-old conjecture here ; a 50-year-old one there . Theoretical computer scientists write proofs for a living, too, so they were bound to put the technology to work as well.
Unclonable encryption exploits a feature of quantum information that classical data do not share: an unknown quantum state cannot be perfectly copied. The goal is to encrypt a message so that someone who gets hold of it cannot split it into two usable versions that reveal the message once the decrypting key is known.
University of Ottawa professor Anne Broadbent and her then student Sébastien Lord introduced a modern framework for unclonable encryption in 2019. Earlier research had shown that unclonable encryption was possible, but existing approaches were either inefficient or needed certain assumptions to be secure. The two new papers claim to show that an efficient version exists without any such caveat.
Ragavan had tried unsuccessfully to solve the problem a couple of years earlier. When he saw it raised again at the Simons talk, he was surprised that it had not been solved. He had also been experimenting with using AI in his own research , so he put GPT-5.6 Sol Ultra to work on it.
Ragavan took a hands-on approach. He directed the Ultra system, which OpenAI says coordinates four AI agents in parallel by default, to work in two-hour stretches. And he checked its progress at each interval and redirected it when necessary. After several rounds, the system produced a proof that he believed was sound, along with a rough draft that he then cleaned up and reorganized.
Ananth and Sahai used the same underlying model differently. Rather than converse with it over successive rounds, they worked off of a bespoke U.C.L.A. system that was designed to help AI models pursue and critique possible solutions. Their paper says the model produced the construction and main proof ideas; the researchers refined and verified the work and say they take responsibility for its claims.
Ananth and Sahai submitted their paper to arXiv.org at 10:35 A.M. PDT. Ragavan submitted his three hours and 18 minutes later. Neither paper was public yet when Lin recognized the overlap and put the researchers in touch. They have since discussed combining the papers into a single version for possible submission to a conference.
Researchers racing toward the same result is not an unusual situation, especially in fast-moving fields. The same AI helping both sides get there is a new one, though.
"This timeline thing is crazy," Ragavan says. "It's like two weeks and a day since this idea even formed."
Ananth was not surprised by the overlap in results. "I was mentally preparing myself that, if we can use ChatGPT to solve this, I mean, everybody has access to it," he says.
The model's first proposed construction also complicated the claim that it had invented something wholly new. "Just before we had to post online," Ananth says, "I remembered, 'I've seen this scheme somewhere.'"
Earlier this year Broadbent and several collaborators published essentially the same construction that Ananth's AI system initially proposed. The new work's contribution was proving that it offered the stronger security that researchers had been seeking.
"In hindsight, this was an obvious thing to look into," Broadbent says. "There's a body of literature, lots of conjectures, lots of schemes. You kind of feel like you gave it on a silver platter."
Broadbent and quantum computing rese...
AIPROPX is an independent multi-source news index — we track, compare, and connect coverage from across the web into one place you won't find anywhere else.