A London-based journalist at The Verge, who specializes in AI coverage and holds the title of Senior Tarbell Fellow, previously reported on health, science, and technology for Forbes.
In a recent discussion on the Cheeky Pint podcast, OpenAI CEO Sam Altman expressed understanding of the challenges faced by mathematicians inundated with AI-generated results over recent months. Despite this empathy, he made it clear that OpenAI has no plans to halt its efforts in this area and hinted at even more discoveries to come.
“I believe humanity has a right to understand the universe,” Altman stated in the podcast episode released on Thursday. “However, I do recognize the difficulties involved in the process.”
Altman’s remarks are among his first acknowledgments of the considerable effect OpenAI's activity has had on the mathematical community, particularly following a series of impressive yet chaotic AI-generated results that have left researchers feeling overwhelmed and struggling to catch up, despite the general enthusiasm surrounding AI's capability as a discovery tool.
During the podcast, Altman argued that the public's right to access new scientific insights takes precedence over the disruptions caused by OpenAI's announcements. He emphasized the importance of disseminating fundamental scientific knowledge and remarked, “We should not withhold major foundational discoveries.”
Many mathematicians and researchers have voiced their frustrations to The Verge regarding the way OpenAI presents its findings. They criticize the company for inadequate attribution, lack of context, and insufficient verification, which they feel forces them into a painstaking process of validation amidst OpenAI's continuous output. This month alone, OpenAI has rolled out hundreds of mathematical outcomes, often accompanied by brief summaries that researchers describe as poorly articulated and difficult to navigate.
Complicating matters, OpenAI is now facing backlash for precisely what Altman suggests it should avoid: withholding potentially critical research. Some in the mathematical and scientific communities have expressed their displeasure with what they perceive as OpenAI’s control over new research, especially after the announcement of three major advancements in mathematics, physics, and computer science. Notably, details regarding one of the major findings related to Yang-Mills theory — foundational to modern particle physics — have yet to be made public. OpenAI has attributed this delay to advice from a select group of experts on the most effective way to release information, a strategy that contradicts earlier guidance from a different group of mathematicians and reflects OpenAI's evolving approach to communication.
During the podcast, Altman hinted that there are numerous mathematical discoveries in OpenAI’s possession yet to be disclosed. When asked if they had more than the hundreds of proofs already shared, he affirmed, “We do. Of course.”
He also admitted that OpenAI's venture into mathematical challenges had at one point “got out of control,” but he considered this an intentional move to test the boundaries of their AI models. He shared anecdotes about an internal Slack channel named “Model Name-Hold My Beer,” where employees were encouraged to tackle the toughest problems they could find.
Altman estimated that OpenAI has invested hundreds of millions of dollars in addressing mathematical problems, with around $10 million (or possibly more) specifically allocated to solving the Navier-Stokes equations, one of the renowned Millennium Prize problems. This expenditure reveals the substantial resources that AI research institutions are applying to scientific inquiries, leading to concerns among researchers that they risk being sidelined in their own fields.
For mathematicians and other researchers looking for a moment of respite, it appears that a pause may not be in sight.
