Post

1. The thing about science that these jokers don't understand is that science cannot be vibe-coded. Whatever its flaws, the point with vibe coding is that you're trying to quickly make something that sorta works, where you can immediately sorta see if it sorta works and then sorta use it.

Carl T. Bergstrom@carlbergstrom.com

“The idea is to put ChatGPT front and center inside software that scientists use to write up their work in much the same way that chatbots are now embedded into popular programming editors. It’s vibe coding, but for science.”

2. Science is a completely different epistemic enterprise. While I don't want to get into the morass of What Science Is, loosely speaking it involves amassing rigorous evidence in support of claims about the natural world.

3. And to do that, you want to *understand* and *justify* every step along the way (at least until you reach one of Latour's Black Boxes) because your end product is a theory, model, or explanation. You can't readily just hit RUN and see if it sorta works.

4. As such, talk doing about vibe science makes no more sense than talk about doing vibe mathematical proofs. As it happens, I get a lot of email from people doing vibe science already. They are convinced that they've discovered the fundamental truths that Darwin and Einstein missed.

26 reposts279 likes

5. ChatGPT has bullshitted them into believing that they're the next genius—but unlike previous generations of crackpot letter writers, they don't even pretend to understand their own "discoveries". They simply tell me that ChatGPT has a proof and they need my help getting the word out to the world.

6. Sure, it's important as a science to stretch your horizons into domains you're not familiar with. But you can't vibe your way through it. This is where either engage in slow months of study or you find collaborators who you trust enough to serve as your guide. A sycophancy machine? Hell no.

I once tried to explain to my father why a Nobel prize in quantum physics did not credentialize someone on global warming. Please let him never discover Claude.

Ew, I have to ask, which Nobel laureate physicist went full climate-dipshit? I'll be surprised if "an expert in Subject A claims to be right about distantly related Subject B, while 99% of Subject B experts are wrong" isn't listed under a diagnosis of narcissism in the DSM-6.

I strongly agree. The trouble I see is that for reviewers is getting difficult to ascertain good science from AI generated slop. I've had a difficult time myself recently, and I had to redo part of the math to see that most of the paper was AI generated gibberish. This may become unsustainable. 😔

That's a weird and pretty bad example since there were a couple successful vibecoded mathematical proofs in the news recently... The key obviously being that formal(type-theoretical) mathematical proofs are easily verifiable whereas general science is not