This week AI (artificial intelligence) has given us bad news. As if the good guy had revealed his dark side to us, filling us with grief. On the one hand, several voices from within those designing and competing to develop AI warn us of the self-destructive danger we are rapidly approaching.
Voices that have reached Anthropic CEO Dario Amodei, who is again calling for a slowdown in AI development itself, to avoid entering the realm of the uncontrolled. Against such an apocalyptic picture, the other source of this week's AI grief seems minor. Yet it plunges us fully into the way AI operates and the logic of exploitation it has set in motion, as it is being implemented.
What a scandal, AI is stealing our ideas!
It is the accusation that OpenAI stole original ideas from mathematicians who had been working for years to offer a solution to the Navier-Stokes problem, one of the great challenges of mathematical science. Now, it seems as if Captain Louis Renault - played by Claude Rains in Casablanca (1942) - has appeared again to say sarcastically with his peculiar cynicism: AI is stealing our ideas! What a scandal!
We all know that AI feeds on the ideas and content of others. Feeding can be considered a euphemism for theft, if such appropriation and incorporation of the ideas and content of others is done without the permission of their creators, nor, of course, any compensation to them. That is what AI does. But here comes the cynical basis: afterwards we all feed on what AI has fed on. So feeding is a little less of a euphemism for stealing.
OpenAI announced this week that one of its AI systems had found a solution to the Navier-Stokes problem, one of the seven Millennium Problems, which had gone unsolved for decades, despite carrying a one-million-dollar prize. An incentive that has driven the work of numerous minds. To do so, according to information from the corporation, it used some ten thousand AI agents for about 88 hours.
Tristan Buckmaster
The brilliant bombshell was quickly clouded. First, because it is clear, according to experts, that OpenAI's solution rests on a mathematical idea previously developed by the Spanish mathematicians Córdoba and Martínez-Zoroa. To that end, they had designed a new strategy - called a vorticity cascade - to tackle the solution of the problem.
However, the one who raised the alarm, with strong accusations, was New York University mathematician Tristan Buckmaster. He too had been working on the problem for some time and maintains that OpenAI behaved unethically by entering a kind of race to publish first and, moreover, potentially using information it had accessed through prior conversations or interactions with OpenAI tools.
Buckmaster states that OpenAI contacted him and his collaborator Levent Alpöge, who worked at Anthropic, and that there were conversations about a possible collaboration. According to Buckmaster, OpenAI wanted him to join its project, but without being able to bring Alpöge with him. Buckmaster rejected the proposal. Shortly afterwards, OpenAI announced its solution.
Human scientists
OpenAI responds that the idea or strategy for solving the problem is one thing and finding the solution is another. It is like saying that the news reports, "features" and other content that AIs deliver daily are different from the content and information they feed on (euphemism) because they give it another touch, a final touch of "personalized summary and articulation".
Now mathematics professionals are discussing what relations they should have with the corporations developing AI and how specifically to use AI. Among other things, they fear a repeat of what happened this week: they talk or publish about what they are working on, still in the development phase, and the AI company takes on the goals and sets its agents running at full speed, to get ahead of the potential findings of human scientists.
The professionals of the main science behind the development of AI thus also become its victims. They have been made to take their own medicine. The logic of exploiting information and creation that has been set in motion with AI, at the hands of mathematicians, affects everyone. Including mathematicians.