0 reactions
I am honestly frightened by how good DeepSeek R1 is at Masters level mathematics
When I was testing ChatGPT about a year ago, I came to the conclusion that AI is pretty good at coming up with solution ideas, but makes some fatal errors when actually executing them.
For ChatGPT, this still holds, though to a far less extent. But for DeepSeek with reasoning enabled, it honestly doesn't hold anymore.
I've been using it for homework help whenever my schedule becomes too busy and I am honestly frightened by the fact that it usually gets a correct solution first try. It doesn't matter how convoluted the arguments get, it always seems to approach problems with a big picture in mind: It's not brute forcing in the slightest. It knows exactly what theorems to consider
The reason it frightens me is that it is honestly far, far better than me, despite the fact that I am about to finish my masters and start a PhD and I have honestly had an easy time, at least in my chosen direction (functional analysis). If that's already the case, will it not only widen the gap and render all but the most ingenious human problem solvers obsolete?
#science
source
When I was testing ChatGPT about a year ago, I came to the conclusion that AI is pretty good at coming up with solution ideas, but makes some fatal errors when actually executing them.
For ChatGPT, this still holds, though to a far less extent. But for DeepSeek with reasoning enabled, it honestly doesn't hold anymore.
I've been using it for homework help whenever my schedule becomes too busy and I am honestly frightened by the fact that it usually gets a correct solution first try. It doesn't matter how convoluted the arguments get, it always seems to approach problems with a big picture in mind: It's not brute forcing in the slightest. It knows exactly what theorems to consider
The reason it frightens me is that it is honestly far, far better than me, despite the fact that I am about to finish my masters and start a PhD and I have honestly had an easy time, at least in my chosen direction (functional analysis). If that's already the case, will it not only widen the gap and render all but the most ingenious human problem solvers obsolete?
#science
source
| 8,000 mlx | |
| 13 | |
| 0 |