14 Followers
120 Following
48 Posts

Studying data science.

Estudando ciência de dados, R e Rust.

Interesses: séries, filmes.

I tried OpenClaw and it's the most insanely bad and dangerous piece of software ever written.

I hope this is peak stupid for AI.

#ai

It would be nice if a Chinese LLM could be announced without the discussion being drowned in sinophobia and casual racism.

#ai

This is a really interesting blog with all sorts of math and computer science things: https://purplesyringa.moe/blog/
purplesyringa's blog

The schema I have in my mind is:

#rust = new C++
#zig = new C
#go = new Python

The most boring discussion on LLMs is the one about ideological alignment of the things.

#AI

@lcamtuf @tqbf he's not wrong, I loved 3B1B and I can't remember a thing from it.

I don't know, in retrospect it makes you feel that you understand enough to keep working on it instead of becoming demotivated. It's important, but a temporary effect.

@vela hey, I love the mastoxiv bots. May I suggest not posting the messages saying that no article was found? Or, if you find them useful for debugging, perhaps use a unique #tag that we can choose to ignore?

Thank you for your great work!

I see many speculative articles on the topic but - in the end, how did LLMs learn to count the number of R's in strawberry?

#llm #airesearch

@j_bertolotti @MarisOzols I've implemented a less-common card game with LLMs and it was the land of a thousand pitfalls.

After all the debugging (of other people's code, which is harder) I'm not sure that I saved a lot of time. Or any time.

It even included some incorrect rules that come from a variant of the game despite my giving it a full specification document that didn't have anything like that.

@simon_brooke @kagihq if you sign up for the plain search engine you only get some basic LLMs that you don't have to use (and if you don't, their companies don't get any money).

I have to say that Kagi Assistant is great though. Customers do want it.