Karpathy sagt, Entwickler hätten eine „KI-Psychose“. Alle anderen sind als nächstes dran.
https://thenewstack.io/karpathy-says-developers-have-ai-psychosis-everyone-else-is-next/
Karpathy sagt, Entwickler hätten eine „KI-Psychose“. Alle anderen sind als nächstes dran.
https://thenewstack.io/karpathy-says-developers-have-ai-psychosis-everyone-else-is-next/
4 Kommentare
The shills are trying to redefine „AI psychosis“ because the current meaning of the term (relating to how validation by AI can hasten mental instability in vulnerable people) invites regulatory scrutiny.
Isn’t pretty much all research, including internal Microsoft research, still showing that AI ultimately doesn’t improve productivity for programmers?
I mean, the nature of the technology is that its limited context windows make it bad at a *lot* of things; but even if that improves substantially, it’s still a technology that is guaranteed to hallucinate and just make shit up.
It’s not just a question of who should take responsibility when AI is used (EG it can’t be fully automated for any serious tasks) — it’s a question of whether it’s fundamentally reliable enough to perform any kind of serious work without needing it to be checked with a fine-toothed comb. And it’s hard to see how that level of quality checking could meaningfully improve productivity when you need a skilled professional doing the work that the AI is replacing. Where is the efficiency gain supposed to be?
There’s some stuff where you just need the work to be „good enough“ but AI outputs can barely even manage that for writing a cover letter, if you can even argue that. I’m just not sure where the real world use cases are here.
Edit: to be clear, I’m by no means against being wrong, it just seems weird to me that the people telling me how great AI is are the same people who don’t seem to be including their time spent reviewing, editing etc AI generated elements of their projects in these massive productivity gains.
There is a huge peception gap with AI and as a developer who has become a yeoman software farmer that doesn’t even use Claude Code, just the Claude chat bot. A lot of developers on reddit completely shit on any comment I make about how we’re kind of trouble in some ways but greatly improved in others.
Like, I can just upload a program in one language and get it a different program language. I’ve uploaded multiple iOS apps I’ve written and get Android apps back. Sure, they’re not perfect, but I don’t even have to debug them and they’re ready to go in an hour.
BUT, I think it’s crazy that companies are laying developers off. I know they’re really doing it because the economy is balls for anything not with „AI“ in the name but blame AI because to do otherwise would call their CEOing skills into question.
Second and most important is that the entire business model of a software company is having fixed labor costs (developers) developing variable revenue streams. You still can’t really make anything notable with Claude, at least, unless you’re a devleoper and have that foundation (and taste!). Laypeople won’t even know what to ask for or be able to navigate Xcode or Android Studio or the entire rest of the system that’s not just the code. A smart company that had a lot of ideas would be hiring developers to get those ideas out but those are never big companies so the developers just get laid off.
There is a LONG and inglorious history of AI hucksters pushing stories of great advances just beyond what investors, business/government clients and consumers can actually *see* or *implement*.
Remember four years ago when that Google employee claimed the AI chatbot Google was developing had become *sentient*? Real media outlets went with that one, as they’ve gone with many other AI hype stories.