we joke that when the AI bubble pops and the managers can't afford the chatbot any more, the surviving companies will hire the people who know how shit works to clean up

[regguy@mstdn.social]

@PurpleStephyr @davidgerard just an FYI corporate bonds are worthless in bankruptcy.

[seanplynch@mastodon.social]

@davidgerard

They'll contract with Accenture and IBM to 'rebiild and modernize' and end up paying double what they were per AI agent (about triple what they were paying per original employee).

[tshirtman@mas.to]

@DBG3D @davidgerard sorry, but i don't think any existing language hasn't been covered much beyond the extent of its actual popularity.

i know jane street, a finance firm famous for being an OCaml house (and making ungodly profit), was disappointed by the support last year or so, and trained their own models, but they ar. very much a special place.

People who want that will have to invent them after the bubble pops.

[fanden@helvede.net]

@ploum "Written in 2022". Shit, have we been in this dumpster fire almost four years already? @aaribaud @davidgerard @audunmb @tante

[kkarhan@c.im]

@davidgerard Consider then a good time to #AbolishCaptialism and #RedistributeWealth in return for not turning #Billiomaires' pronouns into was/were and instead merely sentencing them to lifetime community service without parole…

[bencurthoys@mastodon.social]

@jalefkowit @davidgerard @audunmb @tante Honestly if people stopped making new Javascript frameworks that I'm supposed to have an opinion about and am old-fashioned if I don't use, I would regard that as a massive win.

[gimulnautti@mastodon.green]

@Fulk_It @davidgerard For sure it will be more ”effiecient” to run a codebase that as closely as possible matches both the tokeniser and the training data of the model.

You would need a more expensive model to ”jump over the hoops” of your legacy codebase, that breaks every time the context window is compacted.

[linguacelta@toot.wales]

@jalefkowit

I have some optimism here that we won't actually lose all the programmers, because so many of us do it for the fun and the intellectual challenge. There's little of that in letting an LLM generate bad code on your behalf. So the programmers who are left will be the ones who genuinely enjoy it - probably a lot of hobbyists and open-sourcers alongside industry devs who have enough job security to resist using LLMs (which they know will just get in their way).

[frang@meow.social]

@davidgerard let's see.. skilled contractors will be too expensive.. their previous employees will tell them some variation of "die in a fire"... So.. untrained, unsupervised, entry-level people being paid at as close to minimum wage as they can. They will not have any idea how anything works. Nor will they be able to understand the existing code. So, instead of fixing what is there, they will slather layer upon layer of new (also buggy) code on top of the existing mess to try to patch/correct the problems after the fact.

(Note: I saw exactly this happen back in the 90s.. so there is an existence proof)

[foolishowl@social.coop]

@davidgerard I think the "AI" bubble is one of the more obvious aspects of a long-developing system of social crises coming to a head. "GenAI" is bent on absorbing everything and turning it into slime.

I think there may be opportunities coming out of its inevitable collapse. Part of the difficulty for any radical movement has been that any success gets co-opted or integrated; nothing remains ours. A positive response to the collapse of the "AI" bubble would have to be concerned with things like authenticity and human dignity, with resisting that co-optation.

In cyberpunk, the point is the oppressed survive. We want thriving, not just survival, of course.

There are also the reactionary responses, fantasies of return to an idealized past. Fascism is a species of reaction, that embraces aspects of the new, but older forms of reaction will likely become more prominent.

I imagine decades to come will be filled with grief and profound loss, but some hope if we work towards it.