I’m not necessarily in favor of data centers. That said, every time I see something like this I just wonder when we are going to start hearing about data centers being built in poor countries powered exclusively by coal and zero regulation. I really have no idea what a good outcome looks like but playing ultra hardball seems unwise. I’m very poorly educated on data centers and AI in general. But, I am an expert in electrical generation. I know we could do it pretty clean, at least relative to coal/heavy oil. I don’t think stopping data centers built in the US will magically cease the boulder rolling in the AI direction. Seems like the money is at its back so its going to happen, just a matter of when, where and how dirty.

It’s pretty easy to regulate this:

• Closed loop cooling
• grid upgrade built into design plans, paid for by the datacenter.
• cleaner power generation

I’d say your list is a bit too short. Some more considerations (not comprehensive):

• Construction noise and seismic limits (nearby neighborhoods have been disturbed and experienced damage from blasting operations)
• Operating noise limits (ban on-site gas turbine generation, limit noise levels from cooling towers)
• Limit light pollution

Edit:

• Job protections and guarantees for workers displaced by automation

Still missing heat increase, up to 26°F in the surrounding area. All that heat from the gpus and cpus is a lot. I’ve heard of people using their gpus to heat their apartments.

One square foot of parking lot generates 80 watts of heat in the summer sun. A Walmart parking lot is approximately equivalent to 20 megawatts of heat.

What type of generation are you envisioning here? What type of cooling systems?

My place uses very clean diesel generators. The sound level is below 65 DB with all the generators running - which only happens during an outage. And that’s 10 feet away.

The cooling is fully closed loop. A toilet uses more water a day than the 120,000 sqft DC used in a year.

I work for an AI DC company. We already do all these. We would be happy to have these laws in place - it would make for a level playing field.

Closed loop is absolutely the right answer, and easily regulated.

As to your other two points, the answer is obvious. Nuclear.

It doesn’t need to be nukes. Hydro, solar, wind, and any other mix of power sources is fine, including fossil fuels as an alternative should it be a cloudy, windless week.

Wtf are you talking about? No, any usage of planet-warming compounds for the benefit of creating Twitter incel AI-gen CSAM and TikTok slop videos is unacceptable. Even use of renewables is largely unacceptable, considering the materials (lithium, cobalt, etc) usually entail environmental degradation at best and outright child slavery at worst.

All of those things are foam on the water, the leftover gpu cycles after the bills are paid. No one is spending this kind of money on cat videos.

Our customers are doing biochem/pharma work. Mega-multivariable analysis. There is a big push for a model that can control the innards of a fusion reactor in realtime.

Who said “cat videos”? I said incel slop and CSAM. That’s literally mostly what Grok is used for.

Good for your customers for being the exception and using it for something that matters, but it’s still “fuck AI” in general.

They this on…

And good for you for minimizing your carbon footprint, but it’s still “fuck humans” in general.

We both see that there are large ongoing issues, with bad actors. Elon spending a gazillion dollars for zero cash return is a symptom of his moral destitution, not the usefulness of AI. However painting the world with a broad brush does nothing for improving the world. Bad behavior needs to be punished, while promoting the good.

Leave it to an AI cultist to equate slop-generation carbon footprint with the carbon footprint of actual lifeforms 🤡 Fuck outta here with that bullshit. Fuck your AI.

I pointed to your logical fallacy. Do you see how?

Then you made an ad hominem.

All you’ve demonstrated is your desire to muddy the water.

Would you believe that I see the Luddites had a valid point? The mills were being irresponsible by taking a pure capitalist approach without any social awareness? My company builds more public power infrastructure than it uses, leaving the local community with more electrical capacity than before. Generally by a factor of 2. Why? Because it’s the right thing to do. We don’t build open loop cooling. We don’t build near residential areas. We believe that being irresponsible is not an option. Do we make the news? Of course not, that story doesn’t sell.

aight debate class, you do your thing

I think there are definitely ways to do these data centers that have minimal external impacts, but it costs money and time, and they are trying to rush these through as fast as possible.

We do both. Every day.

Yeah, tough to have a good faith conversation about it.

Some folks dislike AI but use other justifications to hate on them.

Some folks are accelerationists at all costs.

Some read about the criminals in the field (xai) and assume that everyone is like that.

“A person is smart. People are dumb, panicky dangerous animals and you know it. Fifteen hundred years ago everybody knew the Earth was the center of the universe. Five hundred years ago, everybody knew the Earth was flat, and fifteen minutes ago, you knew that humans were alone on this planet. Imagine what you’ll know tomorrow.”

You basically understand what the people with a vested interest in making AI happen want you to know. The truth is that AI is already starting to crumble. It’s a technology that doesn’t do 99% of the things it’s perported to do, and will never do 90% of what they sold it on.

Yeah I’m not versed in the subject enough to say/think you’re wrong necessarily. I do know the general slant Lemmy’s population has against it though.

GenAI as it currently stands is a fancy text predictor. You ever had your phone suggest the next word in a message you’re typing? It’s that, on crack.

When you really wrap your head around the fact that that is all it’s doing, it loses a lot of its appeal imho. Especially for the cost to do so.

To be more specific (for anyone interested), the next word predictors are usually a type of model called an LSTM (at least I think that’s the most common). This model type has been used for a long time for dealing with sequential data. In 2014 there was a famous paper introducing an attention mechanism. This was a rather brilliant, though relatively minor extension to how LSTMs work. Essentially between each step of an LSTM it generates some data representing the model’s knowledge of the sequence to that point. The attention mechanism looks back at these intermediate values and determines how relevant each state is to the current point in the sequence and pulls in the most relevant bits. This vastly improved the memory of the LSTM over longer sequences.

In 2017 there was another famous paper “attention is all you need” which said something to the effect of “the attention mechanism is doing all the work, we don’t need the rest of the LSTM we can replace it by running attention between all point combinations in the sequence.” It’s actually significantly slower to run as the model grows, but much much faster to train because it’s not intrinsically sequential. This is the transformer model that’s the basis of all our LLMs.

Obviously some massive simplifications here but as despite being fairly anti AI, I do love the engineering behind it. So yeah, pretty literally a fancy text predictor, but it turns out when you throw all the compute you can muster at a fancy word predictor is makes the world go crazy

That’s still limited to LLMs. A lot of work being done is not even in any language. It speaks election orbitals and bonding energy. This is used for biomolecular and pharma purposes.

This AI work is normally not classified as “GenAI” which is what I was responding to, and the parent was specifically talking about LLMs. Usually GenAI specifically refers to LLMs as well as image/video creation.

The big data centers aren’t for the most part being built for pharma research.

Most part, no. Significant chunk, certainly. Here are some of the larger ongoing contracts:

​Eli Lilly & Insilico Medicine ($2.75 Billion)

Merck KGaA & Valo Health (Up to $3.0 Billion)

Roche & Manifold Bio (Up to $2.0 Billion)

AbbVie & Gilgamesh Pharmaceuticals (Up to $2.0 Billion)

Novartis & Relation Therapeutics (Up to $1.7 Billion)

Isomorphic Labs (Alphabet) with Eli Lilly & Novartis ($2.9 Billion Combined)

Eli Lilly & NVIDIA ($1 Billion)

Just simply ask yourself, why are all the AI companies discussing going public now? I hope you would agree that AI as it currently stands is far from the human brain replacement it was sold as. Outside of a few very specialized fields it’s basically an email generator. They’re out of training data for all intents and purposes. AI generated content is so ubiquitous now that you can’t use most data moving forward without painstakingly checking it all, and AI is becoming increasingly harder to distinguish cheaply or easily. The widespread adoption has poisoned the well. So AI is as advanced as it’s going to be, and it’s not worth its valuation. They’re all racing for the exit and IPOs are their last hope for their backers to sell and get out before the markets stop being irrational. I hope I’m wrong but that seems to be the writing on the wall.

Edit: they’re also already posturing the current administration for a bailout deal.

Microsoft contradicts your argument on all sides.