Hacker News new | ask | show | jobs
by Tiberium 1183 days ago
> and it crashed

I hope you understand that it's only because of inability of OpenAI's servers to keep up with demand or some issue in their backend code - language models themselves can't "crash" like normal programs on some kind of input, because they "just" generate new tokens.
2 comments
kodah 1183 days ago
I know little more than what I've read about LLMs and other models. I just got it to do them again and it output them rather quickly: https://imgur.com/a/jfviCCe
link
leto_ii 1183 days ago
are the numbers correct?
link
kodah 1183 days ago
I spot checked an random assortment of them and so far yes, but it only takes one wrong one to set every subsequent number off.
link
vineyardmike 1183 days ago
> it only takes one wrong one to set every subsequent number off.

Except its not actually doing that calculation, so one wrong one shouldn't truly affect the rest like "Real" math.

link
mike_d 1183 days ago
> language models themselves can't "crash"

That is like saying the Excel document didn't crash, but Excel did when it tried to parse it. As far as I know there is no proof that you can't cause a LLM to crash with user input.

> because they "just" generate new tokens.

I can write a program that counts to 100 that crashes reliably.

link
inimino 1183 days ago
> As far as I know there is no proof that you can't cause a LLM to crash with user input.

Well, then you simply don't understand how they work.

link
hoseja 1183 days ago
It's a fixed number of matrix multiplications. You can't make matrix multiplication crash by feeding it any sort of strange numbers (unless they haven't handled their Infs and NaNs).
link
mike_d 1182 days ago
> You can't make matrix multiplication crash by feeding it any sort of strange numbers

Matrices don't multiply themselves. You need hardware and software. As I pointed out the LLM is effectively just data that is being processed by a program. It is silly to assume you have no bugs in that software or the underlying operating system.

link
gs17 1181 days ago
The type of bug the first comment was discussing, where recursion could cause it to crash (out of memory? stack overflow?) is categorically impossible based on how the model works, and a bug in the software running it would not change that.
link