sometimes a dragon
he/they, queer, furry, ζ, vegan
“People talk about how much energy it takes to train an AI model. But it also takes a lot of energy to train a human. It takes about 20 years of life — and all the food you consume during that time — before you become smart," the OpenAI CEO told The Indian Express this week.
I would have liked to ask back, how much more food does he require? Gosh, someone offer him an energy bar!
What kind of tasks are on the agenda?
Ugh, I’m so fucking tired of this shit.
I can imagine that an LLM can find bugs. Bugs often follow common patterns, and if anything, an LLM is a pattern matcher, so if you let it run on the whole world of open source code out there, I’m sure it’ll find some stuff, and some of it might be legit issues.
But static code analysis tools have been finding bugs for decades, too. And now that an AI slop machine does it, it’s supposed to bring about dystopian sci-fi alien wars?
Why are people hyped about that?
(Also this poster makes wrong claims about every exploit being worth millions and such, but the rest of it is so much more ridiculous, it drowns out the wrongness of those claims.)
Yesss, and it’s still worth playing today!
Even if you’ve never heard of him before and know nothing else about him… this short tweet alone tells so much about what kind of person he is.
Very impressed with this comment from the creator of the Zig programming language, regarding dealing with AI slop submissions, and generally about LLMs for coding.
I should look into Zig again! Technically, I’ve always leaned more towards Rust, because I like its more uncompromising approach to safety, while Zig always seemed to me a bit more middle-of-the-road on that. But I’ve been disappointed about how wide-spread LLM usage has become in Rust circles, I fear that its culture might tip over in favor of slop. (But it’s not there yet and I hope it won’t happen!)
Anyway, I’m ordering the “Introduction to Zig” book…
This could be regarded as a neat fun hack, if it wasn’t built by appropriating the entire world of open source software while also destroying the planet with obscene energy and resource consumption.
And not only do they do all that… it’s also presented by those who wish this to be the future of all software. But for that, a “neat fun hack” just isn’t enough.
Can LLMs produce software that kinda works? Sure, that’s not new. Just like LLMs can generate books with correct grammar inside, and vaguely about a given theme. But is such a book worth reading? No. And is this compiler worth using? Also no.
(And btw, this approach only works with an existing good compiler as “oracle”. So forget about doing that to create a new compiler for a new language. In addition, there’s certainly no other language with as many compilers as C, providing plenty of material for the training set.)
there isn’t a simple solution to this
How about just not creating the problem in the first place. How about that.
This was very enjoyable! Actually was over too quickly, would have liked to hear you two talk about AI stuff more.
When Woke 2 comes
ooh, please tell me there’s a release date already
Thanks everyone for the replies <3 Guess I should make an account there after all… bleeeh :/
Honest question, since I’m not on linkedin (and kinda looking for a new job): does it really help anyone find a job? It has been my impression from the outside that it’s mostly empty drivel.
I’m confused that anyone thinks that the world needs another linkedin…
Wow. The mental contortion required to come up with that idea is too much for me to think of a sneer.
When all the worst things come together: ransomware probably vibe-coded, discards private key, data never recoverable
During execution, the malware regenerates a new RSA key pair locally, uses the newly generated key material for encryption, and then discards the private key.
Halcyon assesses with moderate confidence that the developers may have used AI-assisted tooling, which could have contributed to this implementation error.
Claim 1: Every regular LLM user is undergoing “AI psychosis”. Every single one of them, no exceptions.
I wouldn’t go as far as using the “AI psychosis” term here, I think there is more than a quantitative difference. One is influence, maybe even manipulation, but the other is a serious mental health condition.
I think that regular interaction with a chatbot will influence a person, just like regular interaction with an actual person does. I don’t believe that’s a weakness of human psychology, but that it’s what allows us to build understanding between people. But LLMs are not people, so whatever this does to the brain long term, I’m sure it’s not good. Time for me to be a total dork and cite an anime quote on human interaction: “I create them as they create me” – except that with LLMs, it actually goes only in one direction… the other direction is controlled by the makers of the chatbots. And they have a bunch of dials to adjust the output style at any time, which is an unsettling prospect.
while atrophying empathy
This possibility is to me actually the scariest part of your post.
Have a quick recovery! It sucks that society has collectively given up on trying to mitigate its spread.
Are you trying to say that you are not regularly thinking about the meta level of evidence convergence procedures?
404 Media: Meta Director of AI Safety Allows AI Agent to Accidentally Delete Her Inbox
This is very serious computing and we must all take it very seriously.