Physiker lösen mit Hilfe von Claude AI ein jahrzehntealtes mathematisches Rätsel: Es „hat meine Sichtweise auf das, was diese Modelle in der theoretischen Physik erreichen können, erheblich verändert“
https://www.livescience.com/physics-mathematics/mathematics/nobel-prize-winning-physicist-and-team-use-claude-ai-to-solve-decades-old-math-puzzle
5 Kommentare
Word on the street is that the physicists were too stupid to be able to fact check it, and Claude just said, „Trust.“
>
And in a total of just 40 prompts, the researchers had a verified publishable analytical solution. To their surprise, this solution was hidden directly within the equations themselves; they didn’t need any external physical assumptions or deep connections between functions.
Interpolation is the thing AI is good at. It disturbs me though that its implied the result was verified by the AI itself?
They also basically admit that if they’d just collaborated with some actual mathematicians they could have probably solved it.
This is a large problem in science in general. Over specialisation and lack of inter field collaboration. So i guess AI can help with that, but its a very resource inefficient solution to what is more an institutional problem.
A year ago I tried all of the LLM’s trying to unite general relativity with quantum mechanics.
Each one said that I found the solution that ties them all together.
I’d be curious to see how many times it’s absolutely dead wrong compared to the chance that it’s right ….nowadays maybe it’s better?
How much fresh water it steal from crops?
The 47% fail rate calculator that says „trust me bro“