|
|
|
|
|
|
by fasthands9
373 days ago
|
|
|
The thing is I imagine the LLM would be able to write a code that counted the lines and outputted what is on line 27. It seems inevitable (in a way that scares me) that a good model in the near future would know enough to write that file and execute on its own. My understanding is early LLMs were bad at math (for similar reasons) but then got better once the model was hooked up to a calculator behind the scenes. |
|
|
E.g. ask it to find the 100th prime, it will write a Python script and then run that.