Luckily, modern cooling techniques like heat pumps are much more efficient than resistive heating, 300% to 500% efficiency is easily attainable. You can keep the inside of a data center nice and chilly at room temperature even in the middle of a heat wave.
Then there is also the heat island effect. Concrete boxes will heat up an area more than dirt or wood, even if those boxes are completely empty inside. If you build them near where people live, people are gonna notice. Houses have gardens, industrial complexes such as data centers and factories don't.
Also, the effect is more concentrated downwind, so the heat effect is not some kind of uniform distribution in all directions.
HN guidelines are that I should choose the most generous interpretation of your comment, not the most likely one. It would be uncharitable and insulting to you to take your comment at face value.
Right, but didn't that one go out the window with your original comment? People follow all the rules some of the time, and some of the rules all the time.
I’m confused. Rejecting one GW of heat to the building exterior with heat exchangers is going to heat up the area around the heat exchangers.
If you had a hypothetical 1 GW generator and used a hypothetical 1 GW loadbank with 1 GW of resistive heating and a fan that was outdoors, the air temperature would rise.
That's surprising to me. You think a gigawatt of heat cannot be felt 500 meters downwind? How near before you experience the temperature rise in the article, if not that?
My wife's hometown in China used to have two options to reach from Shanghai - high speed rail (top speed ~ 200MPH) that takes about six hours or a roughly one hour flight.
Rail travel has stops and slowdowns in more populated areas that make it a bit slower than the advertised top speeds.
These companies aren't coming in, buying property at market rates, and developing it with existing infrastructure under current zoning laws. They usually want tax breaks, major infrastructure changes, and other accommodations and guarantees. It's completely reasonable for people to want a dialog with their representatives before those kinds of arrangements are made with a company on their behalf. And it's entirely reasonable for them to vote out reps that are overly accommodating.
I live in an historic district. I had to attend a public meeting a couple years ago to get approval to change a lamp post. It is perfectly reasonable to ask tech companies to show up and defend massive projects to the public.
Usually the state gives tax breaks and the town gets double it's tax revenue. That's why the councils rush to vote yes, and $20-30M annually is a rounding error for the datacenter.
A meeting to get a lamp post change is exactly how progress stops. It is why we can't build anything in US and it costs a billion dollars per mile of high speed rail.
Communities passed and paid bonds to have water hookusp. Built infrastructure. They should have a say in who gets access and why. There are plenty of places datacenters could build without hassle, but they want access to readymade infra, and that comes with a reasonable tradeoff.
There are also traffic issues--likely less of an issue with datacenters or solar farms. But things like housing that put pressures on local schools are certainly issues--though again not really with datacenters which are more about power and water infrastructure.
Power and water infrastructure being strained can raise utility prices and require new infrastructure, which affect schools and tax rates by extension. It can also lead to new build stoppages until infrastructure is built, which affects hone prices and taxes. There are many knock-on effects.
If it's a brand new type of lamp post that's never been built in that area before and is likely to cause significant problems to the inhabitants (e.g. power blackouts) then that kind of progress should be stopped or at least discussed in an open fashion and not hidden behind NDAs.
Why can the US build data centers that benefit a few rich guys while polluting and poisoning the water for everyone else without any major hurdles, but building high speed rail that the public and the environment would benefit from is nigh-impossible?
In general yes, but a datacenter is to many people a categorically different kind of development because they oppose AI companies. The first datacenter should trigger a lengthy discussion and create rules that apply to all further datacenters.
Yes, rules are absolutely necessary. And community engagement is also essential. Community input might be tiring, frustrating, and the like but people get to speak as part of the process and because they have a right to.
The two are often difficult to dissociate. My town had a fairly fractious town meeting around a rezoning proposal that was mostly for a fairly specific commercial purpose--that passed through a basically procedural mechanism in a second meeting.
in a lot of cases, the leaders of the communities are not following the rules. (see the ppl talking about ndas and such)
in any case, this isn't like "oh we don't want to build an apartment building because it might drop the value of a single family home halfway across town."
it isn't even like "we want to build this train line which will have some negative externalities but the positive effects (and externalities) are worth taking a hit in some areas"
the problems with the datacenters are that like (1) the service its providing (LLMs) has dubious societal value, (2) the direct negative effects such as noise pollution and such have been pretty well documented, (3) the indirect negative effects like massive strain on infra and (4) the people pushing them most heavily are effectively attempting to invade the communities, peddle conspiracy theories about "china" being behind the opposition, and demand to be specially treated because they were bankrolled by big tech, etc.
some people when this topic come up act like anyone opposed is some nimby who hates societal progress or smth and who is super concerned about that their home estimate might go down. but like communities do recognize the need for zoning and restricting certain things being built.
you need the thing being built to both (a) actually be a good that helps the community (or have a very very good reason why some damage to the community is justifiable (datacenter projects generally don't) and (b) need to contain negative externalities (which is why we don't put the chemical plant next to the elementary school even if it's the most economic option). people recognize these things on some level.
There are data centers in my neighborhood and I'd be fine with building more. I live in a dense urban area with lots of buildings of all types relatively close to me.
Yes, I literally support repealing laws that would prevent someone from building an AI data center anywhere it's legal to build any kind of commercial or industrial facility at all. Also as a housing YIMBYist I'm supportive of massive reforms to zoning laws that would result in legalizing many more types of housing construction, including mixed-use zoning laws like what exist in Japan that would make it legal in principle to build housing close to a data center.
> Yes, I literally support repealing laws that would prevent someone from building an AI data center anywhere it's legal to build any kind of commercial or industrial facility at all.
That's not the same thing as the question I asked. That's the answer to "do you support laws that would prevent someone from building an AI data center anywhere it's 'legal' to build any kind of commercial or industrial facility at all."
On the face, not all zoning is the same so it's pretty obvious to me that not all "industrial facilities" are appropriate to build where other "industrial facilities" already exist. For example, there's industrial baking facilities in my neighborhood, but there are not industrial chemical refining facilities in my neighborhood. The latter aren't allowed. By your logic, am I to assume this is okay because it's already the current state of affairs? That just uses an arbitrary date to filter - nonsensical imo. By your logic am I to assume that my neighborhoods prohibition against industrial chemical refining is just inappropriate because we allow industrial baking? We're trying to make sense of the world, not virtue signal via "policy"
Deeper, it's a statement that is so absurdly generalized as to be capable of giving it any useful meaning. You support the construction of AI data centers as a concept? Great. How am I supposed to use this information to navigate any actual disputes about where AI data centers should and shouldn't go.
I'm excited that EVs get to avoid CAFE standards, so we get to have small vehicles again. Even though legacy domestic car manufacturers aren't going to be involved, as they practically wrote the CAFE standards that eliminated small vehicles, and there's bipartisan support to tariff EV imports out of existence, there's still enough market demand for multiple domestic startups.
Unfortunately, we still get abominations like the GMC Hummer EV (~9000 lbs, classified as a class 3 medium duty truck). Rumor has it the model has been cancelled after 2026 luckily.
As long as I get to buy a small vehicle, I don't mind if someone else buys something inefficient. It's better that oversized EVs be on the market than just oversized diesel trucks.
> More than 40,000 people die on America’s roads every year, making them nearly twice as dangerous per mile driven than those in other rich nations. What’s more, the death rate has increased over the past decade, despite the adoption of more sophisticated passive safety systems and advanced driver assistance systems such as lane keep assist and automatic braking. The number of pedestrians killed by motor vehicles has almost doubled since 2010. Why? “Weight is to blame,” The Economist insists.
> Using data for 7.5 million two-vehicle crashes in 14 American states in 2013–2023, The Economist found that for every 10,000 crashes the heaviest vehicles killed 37 people in the other car, compared with 5.7 for cars of a median weight and just 2.6 for the lightest. The publication estimates that if the heaviest 10 percent of vehicles on America’s roads were roughly 1,000 pounds lighter, fatalities in multicar crashes would fall by 12 percent, saving 2,300 lives a year, without compromising the safety of the occupants of the heavier vehicles.
> “For every life the heaviest one percent of SUVs or trucks saves in America,” The Economist wrote, “more than a dozen lives are lost in smaller vehicles. This makes traffic jams an ethics class on wheels.”
> Of course, what’s behind those grim statistics is simple physics. Force equals mass times acceleration: The heavier a vehicle, the harder it impacts an object at any given speed. The core issue in America is the massive differential between the lightest and heaviest vehicles on its roads (and to be clear, The Economist’s reporting is focused on cars, SUVs, and light-duty trucks, not heavy-duty commercial trucks or semis).
> I'm excited that EVs get to avoid CAFE standards, so we get to have small vehicles again.
I don't see evidence for EVs stimulating small vehicle production (but it would be awesome if it were the case). The one smaller vehicle bright spot is perhaps Tesla's Cybercab, which is compared to the Honda Civic, which itself has grown very large over the years. In order to get to that size the passenger mass capacity of the Cybercab is roughly 2/3 of the Civic. [0]
CAFE isn't all they have to deal with. NTSB has stricter safety requirements now and we will probably not have small vehicles again. I just want a damn Hilux.
It could be the case that we’ve reached the last generation of frontier models that can be accessed by the general public. That eliminates a risk that Anthropic could be leapfrogged by a competitor.
Now it’s a competition between products on the near frontier. Anthropic has executed well on products so far. They blew up thanks to Claude Code, not Opus by itself.
I lean libertarian but I can recognize the danger in having access to a machine that can craft pathogens to spec.
A pathogen with a very long incubation time and a high fatality rate would be about as bad as nuclear war. Maybe we need to figure out how to possibly defend against one person doing this before making it easy for anyone to do it.
I’d probably still be on my 13 mini if this feature didn’t exist.
reply