Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

There's a LOT of pushback against the idea that AI is not magic. Imagine if there was a narrative that said, "[compilers|kernels|web browsers] are magic. Even though we have the source code, we don't really know what's going on under the hood."

That's not my problem. That's your problem.



How is this narrative different from the way cryptographic hash functions are thought of? We have the source code, but we cannot understand how to reverse the function. The way modern world functions depends on that assumption.


AI is magic, at times, when is convenient, and it is also extremely scientific and mathy, at times, when it is convenient. Don't you dare doubt those thoughts at the wrong time though.

Just classic hype.


The difference is that neural networks are uninterpretable. You may understand how LLMs in general work, but you can pretty much never know what the individual weights in a given model do.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: