Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The US is winning the AI race where it matters most: commercialization

avkcode.github.io

Post reply on HN
Print view View original

581–590 of 718 posts

Re: The US is winning the AI race where it matters most: commercialization

#581

Post by roenxi »

ben_w wrote: post #568
novok wrote: post #539

Earlier quoted context omitted.

Anthropic is the AI doomer / safetyism lab, and Hinton is one of the patron saints of 'rationalist' AI doomerism. AI doomerism is psychologically attractive to "people with autistic cognitive traits, including dichotomous (black-and-white) thinking, intolerance of uncertainty, and a tendency toward catastrophizing". They are pascal's mugging themselves, to ironically use one of their terms. It's fundamentally a cogni…

I'm reminded of a comic about global warming, "What if it's a big hoax and we create a better world for nothing?": https://climateactionreserve.org/blog/2012/08/31/environment... "What if AI doom is all fear-mongering, and we create AI less prone to make up dangerous stuff or mistake buggy goals for real ones" (which is what alignment is) "for nothing?" Even if Yudkowsky is autistic, you're muddling the condition. Au…

> I'm reminded of a comic about global warming, "What if it's a big hoax and we create a better world for nothing?": https://climateactionreserve.org/blog/2012/08/31/environment...

The people who've made the biggest contribution to creating a better world over the last 50 years have been the Chinese; powered largely by coal and petroleum. And in one of the most ironic results in the 21st century, they're now the leaders in solar panel production on the back of the largest investment in fossil fuel energy in global history.

The comic ran into the same problem as the climate change movement in general - they proposed ideas that generally made people worse off. And if measured in terms of CO2 emissions achieved nothing except pushing wealth creation to Asia. Which, in fairness, is probably appreciated by the Asians.

Re: The US is winning the AI race where it matters most: commercialization

#582

Post by ben_w »

trhway wrote: post #404
dgellow wrote: post #401

Earlier quoted context omitted.

The discourse on this topic is at the point where I have no idea if people are serious or satirical. Please tell me you don’t seriously believe data centers in spaces is a realistic idea

I don't "believe". I'm arithmetically sure that it is going to happen, and it will beat the ground based on pretty much all metrics. Some of my comments with napkin numbers https://news.ycombinator.com/item?id=46882199 https://news.ycombinator.com/item?id=46880680 https://news.ycombinator.com/item?id=46880486 Just a very rough primitive illustration - a land for a house in SV is like a $1M, and putting a 10 ton house…

> Existence of supposedly cheap land somewhere (with not much infrastructure usually) doesn't help as you put your computer nodes into a datacenter building with all the required infrastructure which cost more than the SV land on a sq foot basis.

This is a terrible argument, given that space has zero infrastructure.

Once you can break a data centre into a million sub-units and spread them over a sun-synchronous orbit or ten and cool them radiatively, you can also spread those sub-units on desert land with no water or electricity and cool them radiatively.

The units on the ground would look about 6x larger because ground experiences night and even deserts have clouds, but their PV also lasts 30+ years rather than burning up every 5 years or so, which means the factory making the PV to supply them is the same size.

The main thing you save on is batteries. Tesla already supplies enough batteries that it can manage a "mere" one million 25kW compute modules.

> And that is without consideration of how powerful a weapon is the energy generated by a humongous field of solar panels in space. Remember Reagan's Star Wars? Nuclear explosions as a source of power for the direct energy weapons like lasers, etc. Well, you wouldn't need the nukes anymore. Just redirect a bit of power from your compute nodes. And as i already wrote, the large transnational companies will have to take care about their own defense themselves https://news.ycombinator.com/item?id=47981423 - one more "feudal" aspect of the coming techno-feudalism.

While true, attacking up is easier than attacking down. Anything on the ground has a massive heat-sink all around it, the stuff in space does not. Right now, an attack up is already only limited by the supply of adaptive optics to get through atmospheric distortion.

Re: The US is winning the AI race where it matters most: commercialization

#583

Post by trhway »

ben_w wrote: post #582
trhway wrote: post #404

Earlier quoted context omitted.

I don't "believe". I'm arithmetically sure that it is going to happen, and it will beat the ground based on pretty much all metrics. Some of my comments with napkin numbers https://news.ycombinator.com/item?id=46882199 https://news.ycombinator.com/item?id=46880680 https://news.ycombinator.com/item?id=46880486 Just a very rough primitive illustration - a land for a house in SV is like a $1M, and putting a 10 ton house…

> Existence of supposedly cheap land somewhere (with not much infrastructure usually) doesn't help as you put your computer nodes into a datacenter building with all the required infrastructure which cost more than the SV land on a sq foot basis. This is a terrible argument, given that space has zero infrastructure. Once you can break a data centre into a million sub-units and spread them over a sun-synchronous orbit…

>you can also spread those sub-units on desert land with no water or electricity and cool them radiatively.

no, you can't.

>attacking up is easier than attacking down.

no.

Re: The US is winning the AI race where it matters most: commercialization

#584

Post by lkramer »

chrisco255 wrote: post #443
tenacious_tuna wrote: post #381

Earlier quoted context omitted.

One of my coworkers points out to me every sports reference that pops up in our internal company communications (e.g. "WINNING", "Going to put together a winning team," etc). It seems like everything in the US is couched in competitive language.

Yeah, its no accident that the U.S. is the number one economy, it comes from that kind of thinking across the populace. Complacency gets you conquered.

As a non-American I agree with this. There is a whole different energy to Americans in terms of mindset compared to Europeans (not just in business). I think Europe have outstanding talent, and when it comes together it can be exceptionally good and often in a more sustainable way than the American equivalent, but it's a somewhat sad fact that many of the most successful European companies have been successful by emulating (parts of) the American culture.

Re: The US is winning the AI race where it matters most: commercialization

#585

Post by ben_w »

trhway wrote: post #583
ben_w wrote: post #582

Earlier quoted context omitted.

> Existence of supposedly cheap land somewhere (with not much infrastructure usually) doesn't help as you put your computer nodes into a datacenter building with all the required infrastructure which cost more than the SV land on a sq foot basis. This is a terrible argument, given that space has zero infrastructure. Once you can break a data centre into a million sub-units and spread them over a sun-synchronous orbit…

>you can also spread those sub-units on desert land with no water or electricity and cool them radiatively. no, you can't. >attacking up is easier than attacking down. no.

Asserting the contrary is not an argument.

Nothing prevents SpaceX or anyone else from buying up the right to put these things on cheap desert land. They don't even need to own the land, just the right to wheel these things out on a trailer or a helicopter and leave them there.

A desert is significantly less harsh than space. If your radiator is sized for space, it's overkill in an atmosphere.

And for your edit: https://www.youtube.com/watch?v=xNmbvaUzC8Q

Re: The US is winning the AI race where it matters most: commercialization

#586

Post by ben_w »

roenxi wrote: post #581
ben_w wrote: post #568

Earlier quoted context omitted.

I'm reminded of a comic about global warming, "What if it's a big hoax and we create a better world for nothing?": https://climateactionreserve.org/blog/2012/08/31/environment... "What if AI doom is all fear-mongering, and we create AI less prone to make up dangerous stuff or mistake buggy goals for real ones" (which is what alignment is) "for nothing?" Even if Yudkowsky is autistic, you're muddling the condition. Au…

> I'm reminded of a comic about global warming, "What if it's a big hoax and we create a better world for nothing?": https://climateactionreserve.org/blog/2012/08/31/environment ... The people who've made the biggest contribution to creating a better world over the last 50 years have been the Chinese; powered largely by coal and petroleum. And in one of the most ironic results in the 21st century, they're now the lea…

Mm, there is that.

The unfortunate comparable here is that all the people who care about making sure their AI is safe, regardless of what they mean by that, are beaten to the market by the people who don't.

Re: The US is winning the AI race where it matters most: commercialization

#587

Post by tom2026hn »

kelseyfrog wrote: post #317
Jalad wrote: post #135

Earlier quoted context omitted.

> But if someone invents super intelligence, they can dominate new AI research, control global economies, fight much better, and all very quickly. After reading "If Anyone Builds It, Everyone Dies" I think this is not the correct take. If anyone creates ASI, it just means it's going to wipe everyone out, and it doesn't matter if China or the US do it first

If you were an American, wouldn't you prefer the US wiped you out rather than China?

It would be even better if AGI were to do this.

Re: The US is winning the AI race where it matters most: commercialization

#588

Post by trhway »

ben_w wrote: post #585
trhway wrote: post #583

Earlier quoted context omitted.

>you can also spread those sub-units on desert land with no water or electricity and cool them radiatively. no, you can't. >attacking up is easier than attacking down. no.

Asserting the contrary is not an argument. Nothing prevents SpaceX or anyone else from buying up the right to put these things on cheap desert land. They don't even need to own the land, just the right to wheel these things out on a trailer or a helicopter and leave them there. A desert is significantly less harsh than space. If your radiator is sized for space, it's overkill in an atmosphere. And for your edit: http…

>If your radiator is sized for space, it's overkill in an atmosphere.

no. Again totally wrong.

The 20-40C air surrounding the radiator radiates at the radiator too. This is why a human immediately gets stone cold in space while not in the atmosphere - our body radiates away about 900W and receives 800W+ back from the atmosphere - our internal heat 'generation has to cover only the difference - less than 100W usually.

You probably meant forced convection cooling. That requires additional machinery. And that additional machinery is a significant part why ground based datacenters such expensive to build and operate.

To the comment below:

>The planet underneath anything in low orbit also does this, making this argument irrelevant.

no. Again, totally wrong. You've just stated that a human in LEO wouldn't get immediately cold when exposed to space. Just think about it for a second. And after that plug the numbers in thermodynamic calculator. You'll see your error.

>Likewise, the fact that convection exists even without the adjective "forced".

no. Again, wrong. Non-forced convection is pretty small. Use the calculator. And you'll understand why datacenters use forced convection.

Re: The US is winning the AI race where it matters most: commercialization

#590

Post by ben_w »

trhway wrote: post #588
ben_w wrote: post #585

Earlier quoted context omitted.

Asserting the contrary is not an argument. Nothing prevents SpaceX or anyone else from buying up the right to put these things on cheap desert land. They don't even need to own the land, just the right to wheel these things out on a trailer or a helicopter and leave them there. A desert is significantly less harsh than space. If your radiator is sized for space, it's overkill in an atmosphere. And for your edit: http…

>If your radiator is sized for space, it's overkill in an atmosphere. no. Again totally wrong. The 20-40C air surrounding the radiator radiates at the radiator too. This is why a human immediately gets stone cold in space while not in the atmosphere - our body radiates away about 900W and receives 800W+ back from the atmosphere - our internal heat 'generation has to cover only the difference - less than 100W usually.…

The planet underneath anything in low orbit also does this, making this argument irrelevant. There's even cheap paints specifically made to be most emissive in the wavelength window the atmosphere is mostly transparent to rather than itself emitting at.

As does the fact that humans are only slightly warmer than their surroundings. A human-sized object at the operating temperature of a GPU would have a net radiative loss in Earth's atmosphere of around 0.9-1.3 kW.

Likewise, the fact that convection exists even without the adjective "forced". Again, replace a human with an identically shaped android at maximum GPU operating temperatures of 80-100 °C, normal (non-forced) convection goes from ~117 W (human) to 0.9-1.3 kW (80 °C) to 1.2-2 kW (100 °C).

Post reply on HN