Post B2FjssRCSjxUOSqrGi by okayyeahwhatever@poa.st
(DIR) More posts by okayyeahwhatever@poa.st
(DIR) Post #B2FhAJjbvVss785UVE by WandererUber@poa.st
2026-01-13T19:41:35.683650Z
0 likes, 0 repeats
here's a thought though I'm sure many people had it before. Suppose AI can do a bunch of things and it can do it quite fast too, because you threw compute at the problem.Suppose too that the vast majority of inference subscriptions you can buy right now is subsidized. People theorize this is the case for Claude, chatGPT, grok obviously with it's generous free plan and prompting on X.So what if in the end it turns out it's just cheaper to use a human? How off am I?
(DIR) Post #B2FhMv1h6RPDFNYScS by professionalbigot69@poa.st
2026-01-13T19:43:52.059852Z
0 likes, 0 repeats
@WandererUber >cheaper to use a humanwhere have you been for the past 50 years
(DIR) Post #B2FhUH8XINpFqeGMam by WandererUber@poa.st
2026-01-13T19:45:08.383552Z
0 likes, 0 repeats
@professionalbigot69 I didn't know they had huge server farms doing generalized knowledge tasks in the most roundabout way 50 years ago you knobhead
(DIR) Post #B2FhXnSxYHh6z7Ukgy by professionalbigot69@poa.st
2026-01-13T19:45:49.542710Z
0 likes, 0 repeats
@WandererUber it's always been cheaper to get a human to do anything non-locally
(DIR) Post #B2FhcPnkxoGZ2RQOoa by WandererUber@poa.st
2026-01-13T19:46:38.065386Z
0 likes, 0 repeats
@professionalbigot69 🥚🥚 ✈️
(DIR) Post #B2Fhdq7IbsBcwXFFUe by professionalbigot69@poa.st
2026-01-13T19:46:54.356221Z
0 likes, 0 repeats
@WandererUber hit the oil rigs, nigga
(DIR) Post #B2FheVzcun0MotyiMi by WandererUber@poa.st
2026-01-13T19:47:02.989010Z
0 likes, 0 repeats
@professionalbigot69 I will bro
(DIR) Post #B2FhhDXPzEGOS44Yim by lain@lain.com
2026-01-13T19:47:30.251236Z
1 likes, 1 repeats
@WandererUber it is subsidized right now, although i don't know how much, i suspect many people have subscriptions they just don't use, and that evens out the power users. Even so, if it's not profitable now, it will be in a GPU generation or two.
(DIR) Post #B2FhjpFck2HrzmzdTc by john_darksoul@ariamispainted.world
2026-01-13T19:46:41.909215Z
2 likes, 0 repeats
@WandererUber I asked this question too a couple weeks ago. Monetization of AI is going to be very tricky if it does not get monumentally better than a person. It has a lot of debt to pay back too, so it’s not like there’s a huge amount of time. They sprinted off the cliff and haven’t looked down yet. Some of them orgs (google/fb) have wings. Others not so much. Altmans gambit to starve everyone else of ram seems desperate more than anything. Literally buying time.
(DIR) Post #B2FhtpkDlH9gQE4pFI by WandererUber@poa.st
2026-01-13T19:49:49.065100Z
1 likes, 0 repeats
@lain I know few people that even use chatGPT a bit more than just goofing around with it twice a year and none of those pay for it. One guy, a friend of a friend of a guy I know, has the subscription. His friends all mooch off him until he runs out every month.idk how it is with code though, might be different because if you are the kinda guy that can piss away 200 bucks a month on an LLM then you might not care to maximize usage either.>new GPUtrue, yeah
(DIR) Post #B2FhtqGpo1F83NKs52 by Paultron@shin.mugicha.club
2026-01-13T19:49:48.769447Z
0 likes, 0 repeats
with what a plane?
(DIR) Post #B2Fi8ACyTTmzwYMKLg by WandererUber@poa.st
2026-01-13T19:52:24.161258Z
2 likes, 0 repeats
@john_darksoul interesting.>if it does not get monumentally better than a personYes and I am granting that wrt skill and then thinking about the energy domain here. These models consume over a kilowatt to run, whereas a human brain uses 25 watts in idle and when thinking, and this is excluding hardware and training costs which makes this calculation way worse.Probably some smartass telling me humans are not brains in vats next...
(DIR) Post #B2FjssRCSjxUOSqrGi by okayyeahwhatever@poa.st
2026-01-13T20:12:03.175263Z
0 likes, 0 repeats
@WandererUber What I see is companies are investing in AI IDEs that you can hook directly into your own code base. They’re trying to squeeze something like a 20% increase in productivity and I imagine they plan on there being the downstream benefit of less hiring if your entire engineering department is 20%+ more efficient. From what I see it’s working, although we are in the phase where the AI companies are giving good deals to get buy in. But I imagine they’ll get it to a reasonable price point just by virtue of it being an enterprise level business. I feel like all the home/personal use of AI critiques aren’t really the angle these companies care about, they are going for the corporate world, first by getting their product embedded into the company’s code base, then they’ll slowly bite off more and more - give the AI more control over network controls, auto scaling, once you’re confident in the controls you have in place. As you increase automation you need less people, it seems kind of inevitable to me.
(DIR) Post #B2G1t12mGPRiGAdzCy by GoodBoyUV@poa.st
2026-01-13T23:33:46.006722Z
1 likes, 0 repeats
@WandererUber @john_darksoul Plus humans have to supply their own food, water, etc. so you only pay for labour and overhead whereas you have to supply everything for the machine
(DIR) Post #B2G26oCao0K6UCoOi8 by WandererUber@poa.st
2026-01-13T23:36:16.001194Z
1 likes, 0 repeats
@GoodBoyUV @john_darksoul and if you are a corporation then they come to you from college "hello I have a PhD in Mathematics can I work here?" instead of you having to train a zeroed out spaghetti network simulation from scratch.