How Much Do You Know About Mechanistic Interpretability (AI)?
The way neural networks work is a fascinating subject. The more I learn, the more I realize how little everyone knows about "AI". I have seen many entrepreneurs and even highly educated scientists make many ignorant statements regarding "AI".
The reason I specifically brought up Mechanistic Interpretability is its ability to peak into the inner workings of an "AI". I used to think of them as black boxes with extremely complicated inner workings. It turns out this complexity can be somewhat managed. Currently these techniques work on a single layer. Eventually we could see a multi layered analysis that could unlock previously unexpected features out of the "AI" we build.
0
0
0.000
From what I understand from your post it seems that the potential of Artificial Intelligence is still huge
Posted using STEMGeeks
I watched an interview of the head researcher of the division at Anthropic working on Mechanistic Interpretability. He went into fewer details than the ones in this video you shared, and said Mechanistic Interpretability is new and they are just scratching the surface of what's possible. Basically, from my understanding, Mechanistic Interpretability is kind of similar to analyzing brain activity to see what group of neurons react to what, but for AI, and with a higher probability of better understanding what's going on than in the human brain, for now.