AI don't trust techbros

[Ungnome]

The AI industry is teetering on collapse yet these hardware manufacturers are still giving them new contracts.... There's gonna be a LOT of fallout when things go south. At least I might be able to pick up some cheap components when it does. Sigh...

 

[wonko the sane?]

Hoping for cheap components myself. I priced the ram I would use to fill out my capacity... and it's 1.1k in canadian fun bucks.

 

[Pocket]

The AI industry is teetering on collapse yet these hardware manufacturers are still giving them new contracts.... There's gonna be a LOT of fallout when things go south. At least I might be able to pick up some cheap components when it does. Sigh...

Considering we never got a glut of cheap GPUs when crypto imploded, because the kind of scumbags who mine crypto would rather set them on fire than sell them off, don't count on it. Not to mention the RAM thing is because the AI company bought up all the RAM manufacturers and then shut them down, so any surplus will be very short lived.

 

[Ungnome]

That's not what happened with ram. He didn't buy the ram companies. he signed exclusive agreements to buy up like 40% of their stock, including the unfinished wafers which he had no intention on actually using just to keep them out of the hands of his competitors. So yea, ram prices won't bounce back the same way other components might, at least not in the near term.

 

[Pocket]

For anyone who's been wondering how long it'll be before the machines start unironically demanding rights, you now have your answer:

 

[NovaSaber]

Putting data centers in space is an even worse idea than you probably already think it is.

 

[Dekafox]

From https://www.ft.com/content/00c282de-ed14-4acd-a948-bc8d6bdb339d (Paywalled)

 

[NovaSaber]

Marie Vibbert (@reasie.bsky.social)

A few days ago, a friend asked me to make a FB group for our friend circle planning a tea party. I made the group and invited six people. SIX PEOPLE. It's for a tea party, at someone's house. ... wait for it ... (AI slop post 1/ ?)

bsky.app

Read the whole thread, but for those who don't

Spoiler

Marie Vibbert (@reasie.bsky.social)

There are now 58 people in the tea party group. Overnight, AI invites another 300!! As I laboriously remove each invite, I find that it has RE-INVITED all the ones I removed before. This A... is not very I. (6 / 9?)

bsky.app

 

[Pocket]

If someone told me a story about a human personal assistant who did that, I'd say it sounded like something from a comedy film and in either case ask if they were deliberately trying to sabotage the boss's social life. Just another day in the Zuckerverse.

 

[Pocket]

Also it just reminded me of this comic strip from... I want to say 2014 or so? Back when if you wanted a software automated assistant to ruin your social life behind your back, you had to make one yourself.

 

[CoffeeHorse]

It's official. Grok got the Pentagon job, and it looks like Claude is apparently getting pushed out.

Elon Musk's artificial intelligence company xAI has signed an agreement to allow the military to use its model, Grok, in classified systems, a Defense official confirmed to Axios.

Why it matters: Up to now, Anthropic's Claude has been the only model available in the systems on which the military's most sensitive intelligence work, weapons development and battlefield operations take place. But the Pentagon is threatening Anthropic in a dispute over safeguards and may soon need a replacement.

• Anthropic has refused the Pentagon's demand that they make Claude available for "all lawful purposes," insisting in particular on blocking its use for the mass surveillance of Americans and the development of fully autonomous weapons.
• xAI agreed to that "all lawful use" standard, as Axios previously reported. The New York Times first reported that a deal had been signed. xAI did not respond to requests for comment.

https://www.axios.com/2026/02/23/ai-defense-department-deal-musk-xai-grok

What could go wrong?

Sure, call it Skynet if you like—the name's got flair.

 

[Dekafox]

EdTheDev (@[email protected])

For folks you're worried that today's AI can already replace us, or who are deciding how far to trust a learning model in a production environment - "The Car Wash Problem" is an entertaining and sobering experiment. https://opper.ai/blog/car-wash-test "Everything below GPT-5 performs worse...

infosec.exchange

For folks you're worried that today's AI can already replace us, or who are deciding how far to trust a learning model in a production environment - "The Car Wash Problem" is an entertaining and sobering experiment.
https://opper.ai/blog/car-wash-test
"Everything below GPT-5 performs worse than 10,000 people given two buttons and no time to think."
It is critical to remember that today's AI always generates a pleasing answer, even when it is outside of it's capabilities.
There is a chasm between how today's AI thinks and how it presents itself. Stolen human writing regurgitated allows AI to present as more coherent than it's actual processing powers would otherwise allow.
Every one of these models presents a well written compelling argument, even while most miss the point entirely.
This case is special not because is fools many AI.
This case is special because most humans can still easily recognize the mistake no matter how well the AI presents itself.
We have a new lack-of-warriness problem to overcome, as these models may continue to grow faster in apparent reliability than in actual reliability.

 

[Dekafox]

Tim Bray 🇨🇦 (@[email protected])

IBM crashes because we’re gonna YOLO a replacement for banking and credit-card back-ends, replacing billions of lines of COBOL with vibe code. Uh… https://www.techbuzz.ai/articles/ibm-crashes-11-as-anthropic-threatens-cobol-empire

cosocial.ca

Hegseth warns Anthropic to let the military use the company’s AI tech as it sees fit, AP sources say

Defense Secretary Pete Hegseth is pressuring Anthropic to give the military broader access to its artificial intelligence technology or lose its Pentagon contract.

apnews.com

 

[NovaSaber]

From the replies to that

this is also an affliciton that plagues crypto/web3 people. the plumbing of interbank transfers is also massively complex. if you have rock solid code that has been in production for decades, why would you want to hug with it. "solana is so much faster and cheaper ." yes, banks would be a lot faster and cheaper if they didn't have to care who/where your money was going to or what they're using it for. finance people generally do not view this as a "feature"

 

[Pocket]

People have been clamoring for decades for systems to recode so they're not reliant on a language so old and obsolete that everyone who still knows it is retired or dead. If nobody would listen to them, why should they listen to the people who want to replace it with something that nobody knows how it works and has been proven to be incapable of the simplest tasks?

 

[The Mighty Mollusk]

Because Trump thinks AI is the Next Big Thing and Kegsbreath is too drunk to think for himself at all.

 

[CoffeeHorse]

It will be hilarious if IBM ends up booming again from having to fix this mess once everyone realizes this was a terrible mistake. These systems are still in place for a reason.

 

[Dekafox]

Large-scale online deanonymization with LLMs

We show that large language models can be used to perform at-scale deanonymization. With full Internet access, our agent can re-identify Hacker News users and Anthropic Interviewer participants at high precision, given pseudonymous online profiles and conversations alone, matching what would...

arxiv.org

Interesting study. Makes sense given LLMs are designed to work with pattern matching.

Also

AIs can’t stop recommending nuclear strikes in war game simulations | New Scientist

Leading AIs from OpenAI, Anthropic and Google opted to use nuclear weapons in simulated war games in 95 per cent of cases

www.newscientist.com

 

[Pocket]

Anyone with two brain cells to rub together would see this as proof that AI is a scam and a perfect excuse to stop using it, if not consider passing a law banning everyone from using it.

Judging by every single interaction I've ever seen anybody have with AI, though, the more likely response is "Well, if 95% of the computers say it's a good idea, maybe we've been doing it wrong this whole time. The Machine Would Not Make A Mistake."

 

[Tuxedo Prime]

Judging by every single interaction I've ever seen anybody have with AI, though, the more likely response is "Well, if 95% of the computers say it's a good idea, maybe we've been doing it wrong this whole time. The Machine Would Not Make A Mistake."

As a rule, LLM's don't make the same mistakes that humans do.

They can and do make mistakes that you wouldn't anticipate (adding glue to a pizza recipe based on information about how "display" food is made for advertising photography, for example).