@matrix

1. there's a bubble so yeah in some ways it should slow down
2. these fuckfaces don't want that, they're just trying to legally kneecap free local and open source models before those models kill their businesses

@sun yep, pretty much. they're advertising themselves while trying to kill competition by regulating it out of existence

@sun @matrix if open ai is illegal I'll just still use my illegal models than
@sun @matrix I fucking hate monolithic capital so much. centralize everything! maximize efficiency (only for our business model) minimize resilience (resilience requires delegating power and we don't like that)! make a collapse utterly inevitable and maximize damage when it does!
@sun@shitposter.world @matrix@gameliberty.club With how expensive these things are, the free models are so far behind any of the paid ones and frankly, how productive any AI truly is is still debetable.
These people have already spend gillions on what they can't even meassure as ''10% more productivity''
@Dice @matrix let me rephrase slightly, there's like GLM where I can't run it at home on my macbook, but any very small company with access to a datacenter can start a business with stock GLM (or similar model) fine-tuned to their customer needs, instead of everything being provided by three companies. This is a real threat, and they are scared shitless.

I am currently extremely effectively using ChatGPT and Claude for coding, and running some models locally that are far less capable but I'm getting good utility out of them. They would be useful even if they can't do what ChatGPT can, and running locally is a fantastic luxury.
@sun @Dice @matrix I'm using both a $200 chatgpt and $200 claude subscription right now and keep running out of tokens :sob_smile:
@Dice @matrix the smaller models do always lag greatly in functionality but we're reaching a point where they can do like 80% of the small er stuff that chatgpt etc can do so the small models don't have to be remotely as good as chatgpt to kill chatgpt's business model.
@sun@shitposter.world @Dice@frennet.xyz @matrix@gameliberty.club GLM and kimi are pretty close to claude and chatgpt and with fine tuning would probably be better for certain use cases
@sun @Dice @matrix any recommendations for local models mr moonman? :hapyday:
@sun@shitposter.world @matrix@gameliberty.club I'd be surprised if you're even 10 more productive using them than if you did it like before
Sign in to participate in the conversation
Game Liberty Mastodon

Mainly gaming/nerd instance for people who value free speech. Everyone is welcome.