Put on your math hats for a minute and let’s take a look at what this mid-October problem was all about. It’s a perfect example of what’s happening to AI right now.
Bubeck was excited that GPT-5 seemed to have somehow solved a series of puzzles known as Erdő problems.
Paul Erdős, one of the most prolific mathematicians of the 20th century century, left hundreds of riddles when he died. To help keep track of which ones have been solved, Thomas Bloom, a mathematician at the University of Manchester, UK, created erdosproblems.comwhich lists more than 1,100 problems and notes that around 430 of them have solutions.
When Bubeck celebrated the breakthrough of GPT-5, Bloom was quick to call it. “This is a dramatic misrepresentation,” he wrote on X. Bloom explained that a problem is not necessarily unsolved if this website does not include a solution. That simply means Bloom wasn’t aware of any. There are millions of mathematics articles out there and no one has read them all. But GPT-5 probably does.
It turned out that instead of finding new solutions to 10 unsolved problems, GPT-5 had searched the Internet for 10 existing solutions that Bloom had not seen before. Oops!
There are two conclusions here. One is that incessant claims of breakthroughs should not be made via social media: less reflexes and more gut control.
The second is that GPT-5’s ability to find references to previous work that Bloom didn’t know about is also surprising. The hype overshadowed something that should have been pretty good in itself.
Mathematicians are very interested in using LLMs to track a large number of existing results, François Charton, a research scientist who studies the application of LLMs to mathematics at the artificial intelligence startup Axiom Math, told me when I spoke to him about this Erdős problem.
#social #media #fuels #worst #push
