Water usage goes up with data centers because more cooling is needed when you run the hardware harder.
So: if you're running the models on your own machine, presumably you're not running them as often, and air cooling is sufficient. But, at the same time, this is less efficient in terms of hardware use; the data centers need water cooling specifically because they're getting more bang from their buck from their hardware, by running their hardware harder.
So that's the tradeoff: more hardware-use efficiency means more water usage.
> [Google’s] largest data center, located in Council Bluffs, IA, withdrew an average of 3.9 million gallons of water and consumed 2.8 million gallons per day.
I don't follow. If weights are open, can't competing providers pop up? Including, e.g., coalitions of anarchists who collectively share compute and collaborate on modifications to the weights.
Even if it's too expensive to run the models on your own personal hardware, open weights may still make it possible to take power back from the big private corporations.
I thought Moore's Law came to an end in the last decade?
Certainly the transistors/chip or transistors/$ or flops/$ have not been progressing at the same exponential rate as during 1970-2010. There is still progress, but it's rather slower.
I was careful to say "Moore's law-like". Moore's law is stated as "transistor count per area doubles about every 2 years." [0] As stated, then yes, this might be true but, while important, that's not really the quantity we care about.
As you point out it's really cost per transistor or cost per flop that we mostly care about. I'm finding it hard to find a succinct and clear plot, but I believe one is provided by Our World In Data on "GPU computational performance per dollar" [1] which, to my eyes, clearly shows exponential growth in computational power per dollar.
The picture for storage is a little more muddied but if you squint just right you might still be able to recover an uninterrupted exponential growth [2].
In my view, it's pretty clear that advances in AI have progressed so quickly because GPUs have been keeping up with the exponential growth of computational power (per unit cost).
Exponential growth in this area is usually characterized by "S-curves", where one technology gets saturated but the exponential increase in power or decrease in cost is picked up by another, adjacent, technology, that allows the growth to continue. For compute it's CPUs to GPUs. For storage it's platter drives that are now being overtaken by SSDs.
The more general phenomena is called Wright's law, or experience curve effects [3].
> Practically no one is against hard science research, properly conducted
Eh, I grew up conservative evangelical, and they were pretty much always going to have a problem with research in evolution and astronomy. Same goes for the fossil fuel industry w.r.t. climate science.
When the scientific evidence interferes with religious doctrine or industrial paycheck, then yeah, folks are still going to have a problem with hard science research.
> this is wildly incorrect in an academic context. If I'm using ResNets, I should cite the original ResNet paper, even if I haven't read it.
Eh, I think the correct answer is: read it, then cite it.
You're not really supposed to cite something without reading it, as it might say something different than you think. But sure, citing it w/o reading it is better than not citing it at all.
But then, as you scale towards higher density (e.g., replacing parking lots with garages), you end up with less space for trees. Unless you put them on top of buildings, ofc.
I’d have thought that ‘the free market doesnt make subways’ is a market coordination problem, not an externality one?
Meaning, I think that ticket sales can a good job of capturing the costs and benefits of transportation - the benefits to the consumer and the costs to the producer. It’s worked well in other areas of transportation, like passenger boats and planes, and the market mostly works well in those areas.
So I’d have guessed that issue with subways, different from planes and ships, is that you have to buy the rights to large portions of underground land in order to build your lines in contiguous fashion. It’s hard for private developers to get the rights to these lands, if one landowner refuses to sell, it can block the construction of an entire line. This is why rail, dam, and highway projects tend to be coordinated by the state. They all suffer this problem, which doesn’t apply to transport by air or sea, as air and sea lack similar private property issues to trip over.
> you need to ensure that you cultivate the support of both parties
I'm not sure that that's possible. If someone is motivated by religion or finance to believe something that's simply not true, and your scientific evidence contradicts them, there's a fundamental misalignment. They won't support you, because the real-world evidence contradicts their beliefs, and you're bringing more real-world evidence.
On issues of climate change, evolution, age of the universe, etc., the hard science just straight up disagrees with conservatives. I don't see a way around that.
Alternatively, they've already decided that everyone who works in the field is biased, or close enough, that it's better to just burn it all down.
And, yes, this overlooks the times when conservative scientists like Richard Muller have come in to disprove the mainstream consensus and then come up with the exact same results, but... that's not considered important when they have so many other "examples" of "liberal science".
I know that ultimately, there isn't really any space for 'conservative science' in climate. Either they do bad science, so it's not really 'science', or they do good science and come up with the same 'liberal' answers. But I don't think conservatives have figured this out yet; they're still convinced that good science will prove them right.
So: if you're running the models on your own machine, presumably you're not running them as often, and air cooling is sufficient. But, at the same time, this is less efficient in terms of hardware use; the data centers need water cooling specifically because they're getting more bang from their buck from their hardware, by running their hardware harder.
So that's the tradeoff: more hardware-use efficiency means more water usage.
reply