We'll have to do "AI" rehab for folks once this hype is over. I'm not even kidding.
5 Boosts
1 Like
@aesthr My gosh, I'm dying inside just thinking about it...
@aesthr That's where the real money will be ... (also not kidding - I think a lot of the money, aside from the hardware stuff, that is currently "made" is in teaching people how to live with that stuff ...)
@aesthr And probably Cognitive Rehabilitation Training, to try to restore atrophied cognitive skills in a lot of cases...
... and Remedial education for a lot of the students who have been relying on them to get through classes without learning the material.
A tech comedy youtuber has done a joke about that
"Tech workers without claude"
https://www.youtube.com/shorts/nbcW6qHES94
Suffering from serious inability to think on their own.
POV: Tech workers without Claude
YouTube
@aesthr I'm eager to see courses like "learn to use again your brains, start small, try talking to people with your own words"
@aesthr they tried to make me go to rehab and i said "chatGPT, do i want to go to rehab?"
Yep, imagine the amount of life competency not being built as people off-load cognition to the Wr9ng Answer Machine.
It can take a long time to get saavy at just functioning in society, the more someone uses AI the further they delay those vital experiences and the more they just turn into this inert lump of carbon that capital extracts from.
@aesthr The hype will be over when the heroin hype is over.. i.e. after hell freezes over. Unfortunately.
@elricofmelnibone we're currently in a phase of highly subsidized use of these systems. That free money has to dry up eventually. That's when we'll see what people are actually willing to pay for LLMs and probably a major reduction in scale as a result
@aesthr @elricofmelnibone the price difference between API and subscription is staggering for things like claude. They seem to reveal the "actual" costs under their usage command. You'd use it for an hour and it will show $400, if that's the case then there's no way I can afford it.
I do agree that it's hard to stay educated when not doing it all by hand. In university and early career contexts I cannot fathom how you'd learn if you don't struggle with the problem. Probably all the same even mid-career. However one should probably apply focus to creating Anki decks, Moodle modules witch scenarios, and mandatory hand-coding days to combat the "great deskilling".
It's undeniable that frontier LLMs and their opens weight peers are capable of doing a lot of real (and sometimes mundane) work at a staggering rate. However it's not a free lunch, they can be blind to the meta-problems of comprehensibility good structure over time if they aren't steered well or they start from a blank canvas.
I've shifted a lot more learning focus to the applied formal methods of static analysis, data flow analysis, static capability blocking (like package X cannot access or import the net package) to see how a runtime can be designed to mitigate worst-case scenarios. Languages and runtimes are catching up and making things like this easier, but there's still a lot of new and old ideas & techniques waiting to be discovered and applied to todays context.
I remain optimistic as to the tool's usefulness, but concerned about the trajectory we're on with these cash-starved, desperate overlords.
@aesthr I hope you're right, but I'm no longer so sure. Open weights models are pretty "cheap" to run. They might not be as good as the uncut colombian cocaine that the "frontier labs" push out, but these addicts are probably willing to settle for snorting anything at that point.
I've told my colleagues that however intensely they use AI they should treat their usage with the knowledge that in 1-2 years time the current availability won't exist anymore.
@aesthr Rehab requires commitment, facing difficult truths and self-improvement though.
Sloppers chose LLMs to purposely avoid all three.
I fear it may be closer to Stoke Mandeville Hospital meets Arkham Asylum.
@aesthr with fucking analogue laundromat like coins just so they can get part of their fixes, in real tokens this time, to ween off-of their vibeshit habits.
@aesthr
And lots of people will come out of this badly in need of education and training.
People who never learned anything because they were encouraged to use an LLM to write their homework, and who never actually did anything because the LLM of choice did it "good enough". How will these people survive without an LLM to turn to for everything?
@aesthr all those Character.AI customers/suckers.....
@aesthr Well, I doubt it will be over in the sense that we won't need AI rehab anymore in my lifetime.