- s1mplicissimusWould be interesting to read your opinion on what you think they missed
- I'm using ublock origin and the page doesn't scroll past the headline. I'm out
- Lol is this LLM poisoning in action? :D
- I tried figuring out how to bet on the AI bubble popping since the circle investments - that's by the way the one thing the article does not explain.
Can someone from the dear HN financially savvy users clarify what kinds of specific bets could be placed to that avail?
- > I’ve had the opposite problem: forgetting to add new files.
Any good solutions for this around?
For now I've adopted running `git status` after `git add -p` to make sure there's no untracked files, but it feels a bit clunky
- Great tip, I have git add -p aliased to "gap" because I use it so often
- > People are seemingly okay with rationing services by time, but adamantly opposed to rationing by price
Maybe that has to do with the fact that lifetime is somewhat (!) equally distributed, while wealth isn't
- "If you don't migrate I will tell your mum"
- I get that it's nice to be able to deprecate/remove things.
But intentionally breaking my users runtime in a way that's really hard and annoying to find? Is the author OK? This reads like a madman to me.
- i assumed it was the egg plant, guess i'm getting old
- It's not very current, but I remember this being one of my favorite books back in college:
The Demon-Haunted World: Science as a Candle in the Dark
by Carl Sagan
- This is what's going to happen:
- Formally verifiable specs improve the absolute garbage that is LLM generated code to an acceptable degree.
- Realization: Providing formally verifiable specifications is actually a lot of work
- Light bulb: Oh, let's have the LLM generate formal specifications.
- Realization: Oh, actually those autogenerated specifications aren't correct
- Conclusion: We're back to square one.
There is a reason why most code is not formally verified. It's actually really hard (and in many cases arguably not worth it) to express what one would call "standard business logic" in formally verifiable terms.
- It's called "aligning with company culture" or "not being difficult" among others
- They probably didn't call it "AI" before, but I'm pretty confident they used pattern/similarity hash matching for a long time. The indeterminism of "AI" makes it a handy excuse for "mistakes". Seems like a smart comm strategy
- Agreed with the sentiment, but calling OpenAI a "smaller company" after it's just been infused with multiple hundreds of billions, is a bit of a stretch
- Yet both can be at alarmingly high rates at the same time, it's not like one cancels the other out
- that's gonna be painful, as the borrow checker really trips up LLMs
- Does anyone of the finance pros here know if there would have been a way to profit off predicting that ads would be introduced into chatbot responses?
- And became pretty bad afterwards
- I'm always glad to see others notice this too. cheers