AI winter is in the air! But we think the AI bubble keeps going until 2027 (pivot-to-ai.com)

submitted 7 months ago by dgerard@awful.systems to c/techtakes@awful.systems

32 comments fedilink hide all child comments

https://www.youtube.com/watch?v=vp1-3Ypmr1Y&list=UU9rJrMVgcXTfa8xuMnbhAEA - video
https://pivottoai.libsyn.com/20250823-ai-winter-is-in-the-air-we-still-think-it-has-until-2027 - podcast

all 33 comments

sorted by: hot top new old

[-] froztbyte@awful.systems 11 points 7 months ago

good lord, I barely get to look at the sub for a few weeks and a whole fresh new era of people too short for the ride show up

[-] self@awful.systems 11 points 7 months ago

how did they all choose this weekend to find us? it’s like they’re trying to get their bans in before the downtime

[-] froztbyte@awful.systems 6 points 7 months ago

ADQ any% speedruns

(A = assholes)

[-] BlueMonday1984@awful.systems 9 points 7 months ago

On a wider front, I expect machine learning as a whole will see its funding go down the shitter once the AI bubble bursts - useful ML applications are losing funding in favour of magical chatbots, and the stench of AI is definitely wafting all over it as well.

For an off-the-cuff prediction, I expect the number of AI/ML researchers to steeply drop in the coming years, from funding for AI/ML drying up post bubble, and from researchers viewing the field as too tainted to touch.

[-] dgerard@awful.systems 5 points 7 months ago

oh absolutely. there's a lot of mathematicians who have discovered there's fucking buckets of money right now, and it's all gonna dry up. hope they're socking it away while they can.

[-] DarkCloud@lemmy.world -4 points 7 months ago

It will stick around in the form of general use cases, and specially trained sets.

So Boston Dynamics for instance is training their human robots to move like humans based on specially trained sets for instance.

[-] Architeuthis@awful.systems 28 points 7 months ago

Nobody's using datasets made of copyrighted literature and 4chan to teach robots how to move, what are you even on about.

[-] CorrectAlias@piefed.blahaj.zone 14 points 7 months ago

That example is more ML and not LLM, no?

[+] saltesc@lemmy.world -7 points 7 months ago

Yep. Terrible at many things, very good at others. At the end of the day, very useful technology.

Just as my grandmother always used ti say, "You can't use a knife to beat an egg but you can fuck a player up with it." RIP Gammy. She was the sweetest.

[-] self@awful.systems 21 points 7 months ago

just some uwu itsy bitsy critihype for my favorite worthless fashtech ❤️

how about you and your friend and your grandma all go fuck themselves ❤️

[+] Mirshe@lemmy.world -11 points 7 months ago

There's value in the underlying theory of machine learning. ML models are exceptionally good at sifting enormous amounts of data so long as you're cross checking outputs - which sounds like doing the work anyway, except now you know what to look for and check. Astronomers have been using this shit for years to sift radio telescope data, and particle physicists use it to sift collider results.

Is there danger? Absolutely. But saying it's worthless is like saying the theory of relativity is worthless because it created nukes. Throwing the underlying theory out because people are using it to do shitty things is going to rapidly shrink your world, because a LOT of science has been used to do a LOT of harm, yet we still use it.

[-] blakestacey@awful.systems 17 points 7 months ago

And if we were talking about "the underlying theory of machine learning", you might have a point.

[-] scathliath@lemmy.dbzer0.com 13 points 7 months ago

Most frustrating thing about this AI Psychosis crap, they think machines trained on inherently biased data can communicate fully without bias, and forget the rule parameters of the systems (and that they're products designed in the Skinner mindset of monopolizing the user's time and creating "positive experiences").

The machines aren't designed to scrub bias, they're designed to appear to while alligning with their corporate developer's goals. (Which is also fucked from a consent-autonomy angle if they ever do design AGI, which is essentially what Detroit Become Human was talking about).

[-] self@awful.systems 16 points 7 months ago

you saw this:

LLMs are good for all sorts of things.

and a bunch of waffle about unrelated ML advancements in robotics, and it confused you into giving me a shit lecture on tech I already know about? why?

this post was submitted on 23 Aug 2025

64 points (100.0% liked)

TechTakes

2512 readers

25 users here now

Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.

This is not debate club. Unless it’s amusing debate.

For actually-good tech, you want our NotAwfulTech community

founded 2 years ago

MODERATORS

dgerard@awful.systems