OpenAI Doesn't Understand How Its AI Works

futurism.com

I mean, this is to be expected right, the LLMs are operating at a scale of billions (trillions) of parameters, so it's impossible to explain how every single step or response works.

Overall, people can grasp what an LLM does at a high level. But to debug exactly why an LLM gives a specific response to a specific query, one would have to debug hundreds of billions of parameters.

Hence one class of problems. GPT-5 will have probably an order of magnitude more parameters to debug. If you get more capable systems that "do X" but they didn't interpret X right, and you don't know why they're doing Y to get to X, you didn't even know they could do Y to get X, and they're getting very powerful...

>one would have to debug hundreds of billions of parameters.

That could be unsurmountable, probably requires a computer to do anything of the sort !

Just ask chatgpt to explain its intelligence.

It can't do it well now but who knows, maybe that's the gpt5 breakthrough

Reply on news.ycombinator.com