Merged Artificial Intelligence

[Dr.Sid]

Qwen something or other - give me one second and I'll go and have a look.

ETA: Bloody typical Ollama sid it wanted to do an update and it crashed my system - having to reinstall it! Yeah we really do have to worry about AI taking over!

It already has taken over economy .. but that one will hopefully fix itself

 

[Ziggurat]

Actors like China obviously have to be part of the agreement. The idea is to treat AI like nuclear weapons.

Bwahahahahahaha!

No. China will not be part of the agreement. Even if they claim to be, they won't be. Unlike nuclear weapons, there's basically no way to verify any of it from outside. And China will not let us inside.

 

[Darat]

It already has taken over economy .. but that one will hopefully fix itself

Qwen3.635b and gemma4:26b

The Qwen model is a 20Gb download.

 

[Wudang]

We are so far away from real AI that I can only conclude that the AI firms want discussion directed there to steer discussion towards regulatory capture and away from:
where all the water and power is going to come from,
where all the profits/savings for AI-adopting businesses are
why nobody is using all these AI-written apps
etc

 

[Frank Newgent]

Bwahahahahahaha!

No. China will not be part of the agreement. Even if they claim to be, they won't be. Unlike nuclear weapons, there's basically no way to verify any of it from outside. And China will not let us inside.

The interests of China and the interests of the US certainly ought to be aligned when it comes to the deployment of an AI-assisted weapon of mass destruction, like a virus. It would be very difficult for the CCP to stay in control in the aftermath of such a thing happening.

 

[Andy_Ross]

 

[Beelzebuddy]

The the most in the future: AI is greatly more intelligent than humans. We delegate all powers to it as it's just stupid to decide anything ourselves. AI decides it doesn't need us, or it just ◊◊◊◊◊ up. Yes, even superintelligence can ◊◊◊◊ up. Parents are superintelligent compared to their children. It's not that they can't ◊◊◊◊ up. It's just the children can't see or even understand the nature of the ◊◊◊◊ up.

Closer future: AI is somewhat more intelligent, but vastly faster in decision making. We delegate some powers to it, most likely in crisis, like climate change or war. AI is still very dependent on us, so it won't rationally conclude it doesn't need us. But it can still ◊◊◊◊ up.

Very near future, possibly now: AI is about as intelligent as top 1% of humans. It also has vast knowledge above any human. Bad actors use it to design a weapon of mass destruction, like a virus. Fully synthesizing a virus from scratch is increasingly accessible, and perfectly doable if it's state level bad actors. AI decides nothing, but plays key role.

You're not thinking cynically enough. How about: AI is a spicy chatbot. AI companies have happily been playing VC hype games, throwing BS stories at each other about how their AI is really Skynet in a spicy chatbot costume. That would be fine, except that the leaders of the most powerful country in the world are just about the dumbest sacks of crap you've ever met, are on the losing side of a war they started, and have already fired anyone smarter or more sober than they are. At 3 am one of them takes a break from using AI to draw stink lines on pictures of his rivals to ask if nukes are a good idea, and an evitable tragedy ensues.

Technically that's a failure of late-stage capitalism rather than AI per se, but I think it should count.

 

[theprestige]

What's bugging me lately is that by all accounts, "AI" is an expensive chatbot. And yet all the big players seem quite happy to give away the service for free. They all have websites where you can converse with their chatbots, and they'll happily eat all the enormous costs of those conversations.

I can understand standing up a gigantic new datacenter, that you plan to run at 100% 24/7/365, in order to brute force some complex problem beyond current human capacity to comprehend. I can't understand setting up that same datacenter just to ensure that 8 billion people can get to ChatGPT whenever they want to LARP human interaction.

Unless there's some other reason than inherent value, to enabling these conversations between AI and random humans.

(And no, I don't think "... you are the product" is a complete answer, here.)

 

[The Great Zaganza]

The argument that AI Data Centers do not (yet) use as much power and water as Streaming Data Centers is an argument against Streaming Platforms, not for AI companies.

 

[theprestige]

The argument that AI Data Centers do not (yet) use as much power and water as Streaming Data Centers is an argument against Streaming Platforms, not for AI companies.

Who's making that argument?

Not me, certainly. As far as I can tell, streaming services will never use as much resources as AI. That said, there's plenty of arguments against streaming services, but none of them are relevant here.

---

One big issue is graphics card usage. Even a hard-core, over-clocking, liquid-cooled elite gamer won't be running his GPU at max power 24/7. But for an AI datacenter to be truly cost-efficient, they'll be wanting to run tens of thousands of GPUs at max power, 24/7/365. Not only is that a lot of resources, but it's also a lot of additional wear and tear on the hardware, greatly shortening its lifespan.

Streaming services under typical customer loads are not in the same resource class as AI model training.

 

[Dr.Sid]

Bwahahahahahaha!

No. China will not be part of the agreement. Even if they claim to be, they won't be. Unlike nuclear weapons, there's basically no way to verify any of it from outside. And China will not let us inside.

Of course not, because there will be no agreement.

 

[The Great Zaganza]

none the US will stick to, so why should China.

 

[Wudang]

One big issue is graphics card usage. Even a hard-core, over-clocking, liquid-cooled elite gamer won't be running his GPU at max power 24/7. But for an AI datacenter to be truly cost-efficient, they'll be wanting to run tens of thousands of GPUs at max power, 24/7/365. Not only is that a lot of resources, but it's also a lot of additional wear and tear on the hardware, greatly shortening its lifespan.

The number I've seen used most is 3 years for a GPU in an AI data centre.

 

[Gawdzilla Sama]

View attachment 75838

And that is, perhaps, key. Could the waiter create a pizza out of raw materials? Doubtful.

 

[Wudang]

And that is, perhaps, key. Could the waiter create a pizza out of raw materials? Doubtful.

Does the customer even know what a pizza should be like?

 

[Darat]

The number I've seen used most is 3 years for a GPU in an AI data centre.

And everyone has been quiet about standard defect rates and MTF of all these new GPUs. I'd be surprised if 80% see 3 years out at high usage. Add then another wrinkle in that the next generation nVidia AI GPU is an all new architecture so to upgrade one of these AI datacentres means a complete rip-out and rebuild. And remember these are datacentres being built specifically for AI workloads not general use.

 

[Gawdzilla Sama]

Does the customer even know what a pizza should be like?

LOL. In Sicily the local pizzeria catered to the locals AND to the American airmen on my base, Sigonella, a NATO base.

"American-style" pizza included hard boiled eggs, spinach, peas, and whatever didn't get away from the cat I think.

 

[Andy_Ross]

And that is, perhaps, key. Could the waiter create a pizza out of raw materials? Doubtful.

No but the chef can

 

[Wudang]

On Sicily (Taormina) I ate where the locals did.

 

[Gawdzilla Sama]

No but the chef can

How often do we interface directly with the chef? Enquiring minds want to know.