RE: https://mastodon.social/@glynmoody/116324703722433365

This is a preview of the future. A number of people have already pointed out the obvious parallel between getting people hooked on drugs and getting hooked on using LLMs... it seems we're seeing it play out faster than I'd have expected.

@jzb AI inference is expensive, both financially and energetically. A lot of platforms are masking that by subsidizing usage to chase growth. But that only works for so long: eventually the VC money slows, investors want revenue, prices rise, and limits get stricter.
@Larvitz Yup. So growing a dependence on those tools is, IMNSHO, unwise even if you discount all of the other (many) ethical, social, cognitive, environmental, legal, and technical objections.
@jzb I use a small and open model from France (Ministral) on my own laptop. For surprisingly many repetitive tasks (pattern recognition in files, text redaction), that’s quite good. Without depending on a cloud, without a subscription and without exfiltrating my data.
@Larvitz yeah, that’s a different matter; I’m really just talking about being dependent on a service.
@jzb Anybody who bothers looking up can see the #enshittification at the horizon. Clearly, some people don't want you to look up. #DontLookUp
@jzb thank you but until claude manages to dissolve an ego and turn me into a stone laying still for three hundred years i'm staying with drugs
@lkundrak I'd like to congratulate drugs for winning the war on drugs...