SnozzedandlostMay 23
Elen Le Foll 🇫🇷 🇬🇧 🇩🇪

❝ But I just use AI for boilerplate!" you whimper, clutching your Co-Pilot subscription. Listen to yourself. If you’re writing the same boilerplate every day like some industrial-age cog monkey, automate it yourself. Write a library. Invent a macro. Reclaim some dignity. If AI’s doing your "boring parts", what exactly is left for you to do? Fidget with sliders? Paint by numbers while the inference works it's magic? ❞

I admit I have better things to do this morning than read this, but it's brilliantly funny and soooooo refreshingly on point: https://deplet.ing/the-copilot-delusion/. #LLM #CoPilot #Programming

May 23 at 6:54amWeb
Show threadAndy TonksMay 23
@ElenLeFoll
"It’s autocomplete with a superiority complex"
love people using that description of AI
Show threadVirginicusMay 23
@ElenLeFoll Sheer bloody poetry.
Show threadzedaardvMay 23
@ElenLeFoll That was amazing!
Show threadSebastian BergMay 23
@ElenLeFoll Beautifully written. This could be the beginning of an intriguing novel... 🙂
Show threadVissMay 23
@ElenLeFoll this smells like something @ludicity would dig :D
Show threadLudic 🧛May 25
@Viss @ElenLeFoll If they were not industrial-age cog monkeys they would be able to automate things themselves, but alas, eek eek ook ook
Show threadschrotthaufenMay 26
@ludicity @Viss @ElenLeFoll At least Y.T.’s mom did the coding herself… Smh
Show threadDio9sysMay 23
@ElenLeFoll people are using AI to make boilerplate as if code::blocks didn't write boilerplate for us for years
Show threadJim CarrMay 23
@ElenLeFoll I like to cope using sarcasm too, but my goodness that's a difficult read. It's so crammed with "look how clever I am" analogies, metaphors, and roasts, it's hard to extract the underlying point.
Show threadElen Le Foll 🇫🇷 🇬🇧 🇩🇪May 23
@jcarr I do not think that it was ever meant as a purely informative read. I thought it was highly entertaining, but everyone is entitled to their opinions!
Show threadJim CarrMay 23
@ElenLeFoll Eh, I'm probably being too grumpy about it.
Show threadAnt NewmanMay 23
@jcarr @ElenLeFoll They're not clever enough to know the difference between "its" and "it's" though
Show threadТукс добрийMay 23
@ElenLeFoll A great read, and very on point. 👌
Show threadEric's EdgeMay 23
@ElenLeFoll
“AI isn’t helping you build something novel. It can’t. It only knows what’s been done before. It’s autocomplete with a superiority complex.”
Show threadOsma SuominenMay 23
@ElenLeFoll
Whoa that was awesome! Some of the metaphors were straight from 1980s Gibson cyberpunk. Love it!
Show threadAnneHMay 23
@ElenLeFoll Brilliant!
Show thread[empty]May 23
@ElenLeFoll 'It’s just the ghost of a thousand blog posts and cocky stack-overflow posts whispering, "Hey, I saw this once. With my eyes. Which means it's good code. Let’s deploy it." Then vanishing when the app hits production and the landing gear won’t come down'
Show threadAlbert CardonaMay 23

@ElenLeFoll

"We’ll fill this industry with people who think they’re good, because their bot passed CI. They'll float through, confident, while the real ones - the hungry ones - get chewed up by a system that doesn’t value understanding anymore. Just output. Just tokens per second."

Mutatis mutandis, this is what happened to scientific publishing in academia. Output over understanding.

#ScientificPublishing

Show threadBenjaminHimesMay 24
@albertcardona @ElenLeFoll I’m afraid the tense is wrong here. This *happened* to scientific publishing, and it’s been cemented for several decades now. Long before Alex net
Show threadReinald KirchnerMay 24

@ElenLeFoll I was watching a youtube on music and voice generating LLMs. And one interesting point was, the dude trained an LLM with narrators and improved the model some iterations, and it became real good - way better than the public available models trained on scarped and stolen content.

So I assume, it might be possible to create a real good coding assistent, when you train it on real good code - and not just reddit, slashdot and public github projects. But we are not yet there.