I mean, we’re still at 1 here, there’s a reason it’s outputting Lean code rather than actually doing any computation by itself (Because it can’t do that, it’s a text generator)
No this is just false. Easily provably false but I am not sure it matters here on Lemmy since it seems to be the goal to understand as little as possible then just talk out of your ass with authority.
I don’t think it’s true for everything on Lemmy, but people dislike AI so much on here that it’s true for AI. Even top tier models for now still screw up on some reasoning especially when it involves being reasonable and exercising judgment… for now… But when it comes to very concrete black and white stuff like math or finding programming vulnerabilities, AI is very good right now.
If you ask a local LLM to multiply numbers together without using any external tools, it will still get more right than the average person (who isn’t using any external tools, like pen and paper)
it’s not a high bar because it’s the wrong tool for that, why is that even an argument. Are we evaluating CS students by their ability to compile C code too?
I mean, we’re still at 1 here, there’s a reason it’s outputting Lean code rather than actually doing any computation by itself (Because it can’t do that, it’s a text generator)
No this is just false. Easily provably false but I am not sure it matters here on Lemmy since it seems to be the goal to understand as little as possible then just talk out of your ass with authority.
I don’t think it’s true for everything on Lemmy, but people dislike AI so much on here that it’s true for AI. Even top tier models for now still screw up on some reasoning especially when it involves being reasonable and exercising judgment… for now… But when it comes to very concrete black and white stuff like math or finding programming vulnerabilities, AI is very good right now.
It can do multiplication like you and I can, it’s just not efficient,so why would anyone use it for that.
If you ask a local LLM to multiply numbers together without using any external tools, it will still get more right than the average person (who isn’t using any external tools, like pen and paper)
Thats… Not a high bar, and a low bar for electric sand we fought to think.
it’s not a high bar because it’s the wrong tool for that, why is that even an argument. Are we evaluating CS students by their ability to compile C code too?