500 freaking billions

[TimFox]

Years ago, there were private lending libraries that charged reasonable fees to rent a book.
Big cities like Chicago have large public library systems, including a central library, regional libraries, and branch libraries.
Smaller US cities are not so blessed in the library department.
Some of my former colleagues objected to paying taxes for other people to read library books.

[coppice]

Years ago, there were private lending libraries that charged reasonable fees to rent a book.
Big cities like Chicago have large public library systems, including a central library, regional libraries, and branch libraries.
Smaller US cities are not so blessed in the library department.
Some of my former colleagues objected to paying taxes for other people to read library books.

Yeah, life is complex. Studying people's attraction to subscription vs pay per use models gets interesting. People generally like subscription models, as they like the predictability in their outgoings month by month. Beyond that you might expect high users to prefer subscription models, and low users to prefer pay per use. However, when you do market studies of a large number of people they have wildly inaccurate perceptions of whether they are in the high user or low user category.

[tszaboo]

See  https://www.nrc.gov/reading-rm/basic-ref/students/history-101/too-cheap-to-meter.html
Strauss was suggesting an analogy to city water systems, that often charged a fixed monthly fee rather than charge by the gallon.
(My house's account with the Chicago Water Department is fixed-rate:  I had seen too many stories about billing snafus associated with switching to actual water meters, which is now available as an option.)

Yes, and public transport is either too expensive and in a death spiral because less and less people use it.
Or the ticket sales barely cover the extra costs of the ticket system.

Going back to the original topic, the 500B is even a bigger joke now, that the Chinese released their super cheap AI models.
So the Deepseek was developed on apparently 2048 H100 GPUs, while Grok was made on 100.000. Who would have though that brute forcing everything is not the smartest way of doing things.

[Analog Kid]

Despite Tim Fox's putting a fine twist on the history of the phrase, I still contend that "too cheap to meter" as a cynical response to the optimistic "promise" of nuclear power is the appropriate one. That condition was never to come to pass, as the geniuses who saddled us with that technology never did their homework concerning the costs of the technology, including all the ancillary, hidden costs like those associated with the nuclear-fuel cycle, especially concerning things like waste disposal and plant decommissioning.

Which, BTW, is the real reason most places are hesitant to adopt more nuclear power generation: not so much the result of the agitation of people like me (I was an anti-nuclear activist in a former life), but simply the economic aspects of that technology. It just doesn't pencil out, despite Bill Gate's, et al, efforts to revive it with the "next generation" of reactors.

[SiliconWizard]

Going back to the original topic, the 500B is even a bigger joke now, that the Chinese released their super cheap AI models.
So the Deepseek was developed on apparently 2048 H100 GPUs, while Grok was made on 100.000. Who would have though that brute forcing everything is not the smartest way of doing things.

"Brute forcing" is pretty much the right definition for what we currently do with AI based on large models anyway.
Sure, there are ways to be more efficient, but that's still brute force.
It's a war on data.

[themadhippy]

that the Chinese released their super cheap AI models.

soon to be banned in the usa unless the owners flog it off cheap to a group of american investers

[Analog Kid]

that the Chinese released their super cheap AI models.

soon to be banned in the usa unless the owners flog it off cheap to a group of american investers

It's open source!
Good luck "banning" that.
My hat's off to the Chinese here.

[xrunner]

Nvidia shares sink as Chinese AI app spooks markets

US tech giant Nvidia lost over a sixth of its value after the surging popularity of a Chinese artificial intelligence (AI) app spooked investors in the US and Europe.

DeepSeek, a Chinese AI chatbot reportedly made at a fraction of the cost of its rivals, launched last week but has already become the most downloaded free app in the US.

AI chip giant Nvidia and other tech firms connected to AI, including Microsoft and Google, saw their values tumble on Monday in the wake of DeepSeek's sudden rise.

In a separate development, DeepSeek said on Monday it will temporarily limit registrations because of "large-scale malicious attacks" on its software.

https://www.bbc.com/news/articles/c0qw7z2v1pgo

[Marco]

soon to be banned in the usa unless the owners flog it off cheap to a group of american investers

Multiple US model as a service providers already provide it.

An unfortunate NVIDIA employee also asked for support for model support for their software on their public forum, because they wanted to use it internally for coding and they can't use external models. That post got removed and I assume they got a talking to.

[David Hess]

There are much bigger and more urgent problems with AI and big data.
Companies already using it to maximize misery. How? By using it with dynamic pricing to find the maximum profit possible. And this is on our expense. For example hotels are using dynamic pricing, where they increase the room prices to extract as much money as possible from their customers. And they dynamically adjust the prices through some software, and all hotels are using this software. You cannot affor it anymore, but someone else will (while swearing a lot) and it's a win for the hotel. Why have 5 customers where you can have 4 with more profit?
And this is somehow not illegal, because you see, they don't sit together in one room, discussing how much the price should be, so it's not a cartel. They only do that with some API, so it's legal.
So that's what we using AI for.
Maximizing misery.

If only there was a feedback loop which would attract capital investment in industries where profits are excessive, like building more housing and hotels.  We could call it the economics of a free market.

[David Hess]

Despite Tim Fox's putting a fine twist on the history of the phrase, I still contend that "too cheap to meter" as a cynical response to the optimistic "promise" of nuclear power is the appropriate one. That condition was never to come to pass, as the geniuses who saddled us with that technology never did their homework concerning the costs of the technology, including all the ancillary, hidden costs like those associated with the nuclear-fuel cycle, especially concerning things like waste disposal and plant decommissioning.

In the US, nuclear waste disposal was built into the pricing model from almost the beginning.  Nuclear power is taxed with the funds turned over to the government, which spent the money and left IOUs, like with Social Security and USPS pensions.

https://en.wikipedia.org/wiki/Nuclear_Waste_Policy_Act

Decommissioning is paid for in a similar way.

[tszaboo]

There are much bigger and more urgent problems with AI and big data.
Companies already using it to maximize misery. How? By using it with dynamic pricing to find the maximum profit possible. And this is on our expense. For example hotels are using dynamic pricing, where they increase the room prices to extract as much money as possible from their customers. And they dynamically adjust the prices through some software, and all hotels are using this software. You cannot affor it anymore, but someone else will (while swearing a lot) and it's a win for the hotel. Why have 5 customers where you can have 4 with more profit?
And this is somehow not illegal, because you see, they don't sit together in one room, discussing how much the price should be, so it's not a cartel. They only do that with some API, so it's legal.
So that's what we using AI for.
Maximizing misery.

If only there was a feedback loop which would attract capital investment in industries where profits are excessive, like building more housing and hotels.  We could call it the economics of a free market.

I think on some level more sinister things are happening. For sure someone is already working on AI models, where the feedback loop is people. What kind of information do we have to feed people to change their opinion about something. They already showed that something like a dozen likes are very good predictor on your personality. They could totally do a controlled experiment on just 1% of the user base, and see what sort of info will change your opinion on many things. Like how you vote, what brands you use, what sort of social issues you are interested in. How much you are willing to pay for things. Just do these games on 100.000 people, nobody will notice. Or if they do, "oh it's in beta" and "that's just hallucination".
There are more and more websites and online stores, that only show prices after login.
I really hope it's not all downhill from here.

[David Hess]

There are much bigger and more urgent problems with AI and big data.
Companies already using it to maximize misery. How? By using it with dynamic pricing to find the maximum profit possible. And this is on our expense. For example hotels are using dynamic pricing, where they increase the room prices to extract as much money as possible from their customers. And they dynamically adjust the prices through some software, and all hotels are using this software. You cannot affor it anymore, but someone else will (while swearing a lot) and it's a win for the hotel. Why have 5 customers where you can have 4 with more profit?
And this is somehow not illegal, because you see, they don't sit together in one room, discussing how much the price should be, so it's not a cartel. They only do that with some API, so it's legal.
So that's what we using AI for.
Maximizing misery.

If only there was a feedback loop which would attract capital investment in industries where profits are excessive, like building more housing and hotels.  We could call it the economics of a free market.

I think on some level more sinister things are happening. For sure someone is already working on AI models, where the feedback loop is people. What kind of information do we have to feed people to change their opinion about something. They already showed that something like a dozen likes are very good predictor on your personality. They could totally do a controlled experiment on just 1% of the user base, and see what sort of info will change your opinion on many things. Like how you vote, what brands you use, what sort of social issues you are interested in. How much you are willing to pay for things. Just do these games on 100.000 people, nobody will notice. Or if they do, "oh it's in beta" and "that's just hallucination".
There are more and more websites and online stores, that only show prices after login.
I really hope it's not all downhill from here.

There is a feedback loop like I described in economics, but the lack of a free market explains rising hotel rates, as well as rental and home ownership rates.  Nothing more sinister is required.

AI is being used to create a more efficient market revealing an already existing problem.  If you want lower rates, then increase the supply of hotel rooms, rentals, and housing.  This is a political problem having nothing to do with AI.  Of course the people (special interests) profiting from the current situation will have something to say about that, but the legislators are even more responsible for the situation.  Stop voting for them.

[Analog Kid]

In the US, nuclear waste disposal was built into the pricing model from almost the beginning..

Bullshit.
Provably, demonstrably false.
Up until about the 1960s the industry was still kicking the can of waste disposal down the road. There were not even any usable game plans on what to do with that shit right up until the Nuclear Waste Policy Act was passed in 1982.

I researched this, so I know what the hell I'm talking about here.

And don't even get me started on the Price-Anderson Act, which indemnifies the nuclear industry in the case of any catastrophic incident. Without that prop the industry never would have gotten off the ground in the first place.