OpenAI’s Latest ‘Breakthrough’ Is a Sobering Reality Check

(Bloomberg Opinion) — OpenAI’s path to building super-intelligent machines is looking ever fuzzier.

The latest sign came from an embarrassing blunder by the company’s leading scientists over the weekend. “GPT-5 just found solutions to 10 (!) previously unsolved Erdös problems,” the company’s vice president for science, Kevin Weil, breathlessly posted on X in a post that has since been deleted.

It turned out that wasn’t true. The company’s latest model had simply scraped answers off the internet and regurgitated them as its own.  

Many of the infamous, unsolved problems of mathematician Paul Erdös are listed on this website, and OpenAI’s researchers assumed that no one had figured out the answers to those that were listed on the site as “open.” 

But the website’s administrator, mathematician Thomas Bloom, replied on X that he simply hadn’t updated them on the site. GPT-5 hadn’t solved the problems using first principles. It had just taken the answers from published research papers that Bloom wasn’t aware of. “This is a dramatic misrepresentation,” Bloom tweeted. Weil and the other researchers deleted their posts. 

https://t.co/zc95OyEdPj pic.twitter.com/OmKQh8iH4D— Stefan Schubert (@StefanFSchubert) October 17, 2025

Rival AI labs took the opportunity to needle OpenAI for the mistake: “This is embarrassing,” Google DeepMind Chief Executive Officer Demis Hassabis tweeted. And for good reason. OpenAI has long suggested that the large language models underpinning ChatGPT, which scrape reams of data and synthesize it into clever answers, offer a viable path to artificial general intelligence (AGI). That’s the theoretical threshold where machines become smarter than humans.

The AGI ambition has helped drive sky-high valuations for firms like OpenAI ($500 billion) or Nvidia Corp., which had a $4.5 trillion market capitalization on Monday morning, up from $2.5 trillion this time last year, because it presumes a future where machines will be able to reason and discover the answers to thorny problems in business and society. But the Erdos error is a stark reminder that the large language models underpinning the generative AI boom mostly pretend to be good at reasoning. They are still glorified pattern-matching tools.

Of course that’s useful to consumers and businesses alike. ChatGPT is now being used by roughly 10% of the world’s population and generative AI systems are starting to transform the way many firms conduct research, customer service, marketing campaigns and more. And for many, it has become a near replacement for Google. But the idea that chatbot technology will solve intractable problems in supply chains, managing workforces or product design, still looks to be a distant dream, and not least because it continues to make mistakes.

Tech valuations are soaring thanks to the kind of puffery we saw this past weekend, where OpenAI’s lead scientists rushed to make an announcement before checking the facts. History shows that overpromising has a tendency to slow progress, and past AI winters were sparked by overinflated expectations, not failures. As I’ve argued before, substantial breakthroughs on reasoning will probably require approaches to machine learning that aren’t as fashionable as generative AI, such as neurosymbolic AI. 

Genuine reasoning that pushes the frontier of scientific knowledge shouldn’t be confused with ChatGPT’s ability to remix the world’s knowledge, especially by the company’s own researchers. But don’t expect that to stop Silicon Valley from telling us that superintelligence is just around the corner, or from the AI boom continuing to run on belief as much as it does on genuine breakthroughs.

More From Bloomberg Opinion:

This column reflects the personal views of the author and does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.

Parmy Olson is a Bloomberg Opinion columnist covering technology. A former reporter for the Wall Street Journal and Forbes, she is author of “Supremacy: AI, ChatGPT and the Race That Will Change the World.”

More stories like this are available on bloomberg.com/opinion