Just a few more bucks bro! I swear then it will be the revolutionary "AI" we promised it to be.
this post was submitted on 19 Aug 2025
213 points (94.6% liked)
Technology
74233 readers
4393 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 2 years ago
MODERATORS
Nah, it's good that they ripped off that bandaid. Parasocial AI relationships are terrible.
Happy cake day!
we definitely need to eradicate tech ceos from existence
How about your responsibility for the damaging and lethal product of yours, OpenAI?
Eh. Your load of money made a oopsie. Another load of money will surely fix it.
It annoys me that Chat GPT flat out lies to you when it doesn’t know the answer, and doesn’t have any system in place to admit it isn’t sure about something. It just makes it up and tells you like it’s fact.
LLMs don't have any awareness of their internal state, so there's no way for them to see something as a gap of knowledge.
Took me ages to understand this. I'd thought "If an AI doesn't know something, why not just say so?“
The answer is: that wouldn't make sense because an LLM doesn't know ANYTHING
Wouldn't it make sense for an ai to provide a confidence level though?
I've got 3 million bits of info on this topic but only 4 of them lead to this solution. Confidence level =1.5%
It doesn't have "3 million bits of info" on a specific topic, or even if it did, it wouldn't be able to directly measure it. It's worth reading a bit about how LLMs work behind the hood, because although somewhat dense if you're new to the concepts, you come out knowing a lot more about what to expect when using them, what the limitations actually are and how to use them better if you decide to go that route.
It's a feature. Not a bug of LLMs.
It doesn‘t know that it doesn‘t know because it doesn‘t actually know anything. Most models are trained on posts from the internet like this one where people rarely ever just chime in to admit they don‘t have an answer anyway. If you don‘t know something you either silently search the web for an answer or ask.
So since users are the ones asking ChatGPT, the LLM mimics the role of a person that knows the answer. It only makes sense AI is a „confidently wrong“ powerhouse.
It wouldnt finish a lyric for me yesterday because it was copyrighted. I sid it was public domain and it said "You are absolutely right, given its release date it is under copyright protection"
Wtf
yeah, there are guardrails but for copyright, not for bullshit. ig they think copyrighted content is worse than bullshit.
Chat GPT makes up everything it says. It’s just good at guessing and bullshitting.
It's literally a guess machine ..
It doesn't admit anything, it's a language machine
In the end it's a word generator that has been trained so much it uses facts often enough to be convincing. That's its basic architecture.
You can ask it to give a confidence level to have an indication of how sure it is of the answer.
Chat GPT makes up everything it says. It’s just good at guessing and bullshitting.
Someone I know (not close enough to even call an "internet friend") formed a sadistic bond with chatGPT and will force it to apologize and admit being stupid or something like that when he didn't get the answer he's looking for.
I guess that's better than doing it to a person I suppose.
The objections are about its personality? Who cares, as long as it's good at coding? That's the only thing it's actually useful for.
Many people wanna fuck their AI
Well one thing's for sure, data centers are going to be insanely cheap in the near future.
And they'll all be optimized for GPU workloads :(
If anyone actually spent money on science anymore, I bet this would be great for, like, protein folding, that sort of thing.
Terrible for running websites though.
that’s actually okay… the only thing that’s different about GPU workloads is that they’ve very energy dense… as CPUs and other hardware progress, their energy requirements get more dense… 10 years in the future, today’s GPU optimised datacentres will be perfect for standard workloads
… unless they’re centrally liquid cooling the whole DC, which i’ve heard discussed but is a very new concept with a lot of unknowns
GPUs are only good for workloads that multi-thread really, really well. That's why we don't just use them as CPUs.
The idea that today's GPU will be tomorrow's CPU makes no sense. We've had GPUs for ages. If they were capable of being used in place of CPUs we'd already be doing it. Why aren't yesterday's GPUs today's CPUs?
yes, but we’re talking about hardware requirements… data centres aren’t really designed for the software that runs in them; they’re designed for the hardware… a “GPU optimised” data centre just has a lot more power running to each cabinet, and has to have a lot larger cooling capacity in a small area
the hardware inside the data centre can be swapped out: it’s not like GPUs are built into the foundation of the building
OK, if we're talking about infrastructure rather than specific equipment, then yes, I would broadly agree that the datacentre infrastructure itself can be repurposed.
Unfortunately, by that point the whole data centre will already have been sold off for parts because its never going to recoup its initial investment in the first place, and throwing even more money into swapping out those GPUs for CPUs is going to be a complete no go.
yes. the comment was
Well one thing's for sure, data centers are going to be insanely cheap in the near future.
which i think broadly agrees with your thinking… the hardware will be sold, but the building and utilities will remain… thus, data centres will be cheap to buy and repurpose as AI companies try and offload them… might possibly see some cheap AF colo or dedicated options in the future
“I literally lost my only friend overnight with no warning,” one person posted on Reddit
It was meant to be satirical at the time, but maybe Futurama wasn't entirely off the mark. That Redditor isn't quite at that level, but it's still probably not healthy to form an emotional attachment to the Markov chain equivalent of a sycophantic yes-man.
Markov chain equivalent of a sycophantic yes-man.
not only that, but one that is fully owned and operated by a business that could change it any time they want, or even cease to exist completely.
This isn’t like a game where you could run your own server if you’re a big enough fan. if chatgpt stops existing in its current form that’s it.
After reading about the ELIZA effect, I both learned how people are super susceptible to this, and just need to remember the core tenants of it to avoid getting affected:
There's an entire active subreddit for people who have a "romantic relationship" with AI. It's terrifying.
load more comments
(3 replies)
load more comments
(7 replies)
"we fucked up our massive new generation product launch.. oh well lets invest trillions in new data centers" How do investors keep falling for this shit.
He's saying the launch was done badly because some users are in love with GPT-4 and it should not be removed. From a point of view of a investor having people addicted to your product is a good thing.
How do investors keep falling for this shit.
The ROI and the supposed savings from getting rid of the human side of technical support but also efforts of human creatives.
load more comments
(7 replies)
view more: next ›