Could using AI to help you with your next big idea end up putting AI giants ahead of you?
A professor and OpenAI exchange views on progress on a math problem so difficult to solve that a $1 million prize has been offered for it. The whole saga raises questions credit for AI-powered work and the data used to train modelswrites Ben Shimkus of BI.
NYU math professor Tristan Buckmaster announced major progress on the Navier-Stokes problem, only to be followed less than 24 hours later by OpenAI, which proposed a full-fledged solution.
(Even if you skipped Algebra in high school, stick with me here. I’ll keep it at a high level.)
Buckmaster was the first to report the progress he had made in tackling the problem using AI models, including those from OpenAI. His post also described a tense back-and-forth with OpenAI over their competing research.
According to Buckmaster, OpenAI suggested coordinating all announcements and not mentioning Buckmaster’s partner, Levent Alpöge, who works for (wait for it) Anthropic. (You can read Buckmaster’s entire post here.)
OpenAI computer scientist Sebastien Bubeck pushed back against parts of Buckmaster’s post, saying he “never asked Levent to be relieved of authorship of his own work.”
Buckmaster said OpenAI acknowledged it was working on the problem after it learned of his investigation, but told him that “the model had not looked up any user data.”
OpenAI on Tuesday published what they claim is a solution to the Navier-Stokes problem. The company said it had not seen Buckmaster and Alpöge’s work before they released it publicly, but also could not rule out that “anonymized data derived from their use of our products helped improve our models.”
You don’t need to understand long division to see some potential problems here.
Does using AI models mainly mean running the risk of their owners putting your idea first?
According to OpenAI, the evidence is very different from the work of Buckmaster and Alpöge. And Buckmaster didn’t accuse OpenAI of anything specifically, saying he just wanted to outline the timeline of events.
But let’s say you had a great idea for bridesmaid dresses that you’re playing with in ChatGPT. Just before you’re ready to launch, OpenAI announces it’s releasing its own app focused on bridesmaid dresses. Coincidence… or conspiracy?
(Complex math problems are much closer to OpenAI’s wheelhouse than bridesmaid dresses, but you get the idea.)
This problem is not limited to academia and hypotheses. From Apple trade secrets lawsuit against Open AI expressed similar concerns. When a company’s secret sauce is shared with AI, it can “create irreversible and perpetually propagandistic use of the trade secret.”
OpenAI also discussed ways to capitalize on the ideas of users who are becoming major companies, in part thanks to the technology. CFO Sarah Friar pitched OpenAI to get paid as customers AI-enabled work makes money. Friar used a pharmaceutical partner as an example: When OpenAI technology helps develop a breakthrough drug, it can take a licensed portion of the drug’s sales.
When everyone is on the same page, it can be a great symbiotic partnership. But I bet most of us aren’t fully aware of the terms and conditions of all the AI apps we dump our ideas into.
