Merged Artificial Intelligence

[Shepherd]

Zooming in on some of the red dots, they look to be in places I would expect them to be. There doesn't seem to be much information yet, though, on who they would belong to, whether schools or businesses.

ETA: ...or government labs.

 

[steenkh]

I've been following him since earlier in the year when I started to try and find information on the proposed and promised massive new datacentres and I couldn't find any that were finished and many had not even broken ground. At the moment he is talking a lot of sense (I know this because he agrees with me).

In Denmark this is seen as a real problem because the Danish energy net operator EnergiNet is handling energy requests on a first come first served basis, and that means that because of all the so-called ghost data centres that will likely never be built, and which have applied for a large share of the available electricity, common people, and companies who want to convert to EVs can't get charging stations.

There is now a new political movement towards putting data centres at the back of the queue.

 

[dirtywick]

Palantir CEO Calls Drugs 'The One Thing I Don't Do' in Wildly Unhinged Interview

"This is the voice of American business that is being channeled through me," Alex Karp declared on national television.

gizmodo.com

palantir ceo alex karp goes on news with rant about the ai bubble, insists he's not on drugs

“Okay, so that’s… that’s my role,” said Karp this morning, sort of getting back to the topic of AI. “But I’m telling you, in this country, at every single enterprise I deal with, these people are livid. They’re like, ‘I am paying for tokens that create no value.'”

Karp also seemed to rail against the proposals for a kind of tax on AI companies that would help fund jobs programs or pay a dividend directly to citizens, something that Anthropic has floated as an idea if unemployment rises to 25%.

 

[Gawdzilla Sama]

I'm glad I'm old.

 

[Wudang]

404Media say they have emails showing how companies are worried about AI costs

Companies Are Throttling Employees’ AI Use Because It’s Too Expensive

Sources and leaks from Amazon, Adobe, Atlassian, Citi, and more show what is really happening with AI right now: companies are trying to rein in AI use as costs spiral out of control.

www.404media.co

 

[Shepherd]

I just paid last month's electric bill and noticed that it was significantly higher than this time last year!

I've been running some local diffusion models just to mess around with them, both image and video, using my two 150 W cpus and my 75 W gpu. So, apparently running at approximately half a kilowatt for hours a day kind of adds up! Who would have thunk it?

In the wintertime it's just a part of my heating bill, but this time of year, it's really hard for me to justify this. Around now and the next few months are typically my lowest electric bills of the year.

So, as neat as they look, I don't think my dancing humanoid robots are worth the cost.

 

[Dr.Sid]

I just paid last month's electric bill and noticed that it was significantly higher than this time last year!

I've been running some local diffusion models just to mess around with them, both image and video, using my two 150 W cpus and my 75 W gpu. So, apparently running at approximately half a kilowatt for hours a day kind of adds up! Who would have thunk it?

In the wintertime it's just a part of my heating bill, but this time of year, it's really hard for me to justify this. Around now and the next few months are typically my lowest electric bills of the year.

So, as neat as they look, I don't think my dancing humanoid robots are worth the cost.

Oh, mine totally are.

 

[Tero]

I can see where this AI thing is going in the long run.

Asking AI a question:
-thinking
-still thinking
-what's the hold up?
-need one more minute for the solar panels to generate the power for your question
- so it's the carbon footprint?
-yes
-I could send you a few lumps of coal for next time
-that would help

Goes on line to buy virtual coal and ship it to Iowa.

 

[Gawdzilla Sama]

I just paid last month's electric bill and noticed that it was significantly higher than this time last year!

I've been running some local diffusion models just to mess around with them, both image and video, using my two 150 W cpus and my 75 W gpu. So, apparently running at approximately half a kilowatt for hours a day kind of adds up! Who would have thunk it?

In the wintertime it's just a part of my heating bill, but this time of year, it's really hard for me to justify this. Around now and the next few months are typically my lowest electric bills of the year.

So, as neat as they look, I don't think my dancing humanoid robots are worth the cost.

Our electric bill last month was double the same month last year. Triple digit temps will do that to a feller.

 

[Puppycow]

This is a channel about math (or maths as the Brits call it). It's about AI models solving math problems (Erdős problems) that had not been solved yet. Details and links to relevant information are in the video description on YouTube.

This next one is just a basic explanation of why AI is so expensive to operate (expensive literally in the amount of electricity consumed). It's maybe longer than necessary, but the guy knows what he is talking about. If you want to know exactly what a token is and how AI models use them, this has it.


Both of these channels are good, reputable channels, and not slop factories.

 

[Gulliver Foyle]

I can see where this AI thing is going in the long run.

Asking AI a question:
-thinking
-still thinking
-what's the hold up?
-need one more minute for the solar panels to generate the power for your question
- so it's the carbon footprint?
-yes
-I could send you a few lumps of coal for next time
-that would help

Goes on line to buy virtual coal and ship it to Iowa.

The new big thing in LLM apps is loops, whereby you set an LLM "agent" to do a task like coding software and it loops through the steps continuously until it's told to stop.

Looks to me like it's a panic reaction to the massive constraints customer companies put on staff using LLM's once Anthropic and Open AI started charging at cost (well except for all the free tokens they're now giving out to entice peoe back).

 

[The Atheist]

Very good piece from Ars Technica on one of the imminent dangers of AI: https://arstechnica.com/security/20...popular-ai-tools-to-assemble-massive-botnets/

 

[Dr.Sid]

IMHO it will eventually shift to local .. sure, at the moment local models are still pretty much useless. But they are imprivn

Very good piece from Ars Technica on one of the imminent dangers of AI: https://arstechnica.com/security/20...popular-ai-tools-to-assemble-massive-botnets/

Priceless !

 

[Frank Newgent]

IMHO it will eventually shift to local .. sure, at the moment local models are still pretty much useless. But they are imprivn

Priceless !

Into the valley of steel.

 

[Trebuchet]

My brother works for a convenience store chain. The employees were required to watch a video of a woman explaining to them what a terrible thing it would be for them to unionize. He snapped a pic and showed it to me.
Ok...wait... is she..... Yup, he said. AI. Quite deep into the uncanny valley.

 

[Emily's Cat]

Your first quote let the cat out of the bag, it admits that LLM "reasoning" gets things right by chance.

Nah, it gets things right by “lots of people say…”. Ultimately, AI is almost entirely argumentum ad populum. Most of the time, most people within a given culture agree when it comes to moral questions, societal norms, and values. Most of the time, most people agree about facts.

But sometimes, most people are entirely wrong or subject to propaganda or misinformation or ignorance.

 

[SpuriusMaximus]

A 2026 study coined the term "trendslop" for it. It's not even a case of most people agreeing. It just gives you whatever advice is popular at the moment online, such as social media, which is probably the worst place to look for any competence in who agrees with what. And it's even worse for business advice, which is what the study was about. It gives you the same buzzword ridden advice no matter what parameters you actually gave it about your business.

 

[SpuriusMaximus]

Also, speaking of AI failures, dunno if it's been said before, but AI is increasingly getting used by recruiters and/or HR to assess some candidate's competence. Well, it turned out that you can trivially game it by just changing your Zoom background to a bookcase. The same person, saying the same things, was rated as more competent if they had a bookcase behind them.

 

[The Great Zaganza]

the standard AI doomsday story is the Paperclip Scenario.
Turns out you don't need an AI for that - humans will happily turn everything into compuitronium at any price rather than admit that they bet on the wrong horse.

 

[SpuriusMaximus]

404Media say they have emails showing how companies are worried about AI costs

Companies Are Throttling Employees’ AI Use Because It’s Too Expensive

Sources and leaks from Amazon, Adobe, Atlassian, Citi, and more show what is really happening with AI right now: companies are trying to rein in AI use as costs spiral out of control.

www.404media.co

I think the even bigger problem is not just being expensive, it's that you don't know how expensive it will be. You don't know in advance how many tokens a certain task will need to complete, and even less so after the several tries and tweaks to the prompt. You only know how much you've used when the invoice comes.