Yeah, we’ve been through this exact same game with multiple iterations of Intel and AMD chips. When AMD first started doing consumer CPUs they badged them according to their equivalent Intel clock speed because one to one comparisons were misleading.

What’s the L1 and L2 cache? What are the bus speeds? How many cores and how are they architectured? Multi-threading? How many steps is the instruction cycle? There are so many factors beyond just clock speed that play into real world performance.

I think that’s 100% what this is, and it’s a very smart play if that’s the case. Intel are reeling from some significant setbacks, while Nvidia is swimming in cash. There’s never been a better time for them to make a play for the desktop CPU space.

And they’ve got absolutely no illusions about what’s happening with AI. They’re the ones who are literally paying AI companies to buy their chips. They know the space is collapsing. But as the guys selling the picks and shovels, they can ride out that collapse if they’re smart.

End of the day, if what we get out of this is a new, serious competitor in the CPU space, that’ll at least be some kind of win. With Nvidia’s money and expertise they could really force Intel to get their shit together. AMD chasing their heels is the only that’s ever kept them from completely going to shit, but more competition is even better. With all three major companies playing in both the CPU and GPU spaces, that could be really good for consumers.

It annoys me that I can guarantee they’re doing this to make it easier for agentic AI to interact with the web, but I guess if we get some important accessibility benefits along the way that’s not a bad thing.

Yes, the bottom of the ocean is a terrible place to put a data centre. And the fact that it is, somehow, still a more practical option than space is a really good indicator of how unbelievably stupid the entire notion of space data centres is.

Seems reasonable. This case is substantially similar to previous cases that were taken up by the supreme court - in particular a finding over whether a selfie generated by a monkey was copyrightable - and the lower court decisions are in line with the previous precedents set by the supreme court. So they’re effectively just saying “Our opinion hasn’t changed.”

Yep. Radiation is deadly to computers, and without the atmosphere to protect you there is a LOT of radiation in space.

Basically the way you would make a stealth spaceship would be by focusing as much as possible on energy efficiency. At every juncture you would try to use as little power as possible, and use every bit of it as efficiently as possible, so that you’re not remitting waste. That waste, in the form of heat, radio waves, etc, is what gets you spotted.

You could also run heatsinks temporarily for enhanced stealth as you suggest, then open up radiators to cool them - or eject them - once it’s safe to do so.

(For the Elite: Dangerous players, yes, that game got it right.)

The entire ISS has 14GW of cooling (and a lot of that just goes towards keeping the sun from cooking it). A single server rack can produce around 72GW of heat.

The ISS cost about $100 billion.

Basically, if you took the entire budget of Sam Altman’s “Stargate” project (money that, to be clear, he does not have and will not get) and put it into space data centres you might, optimistically, put one rack in space.

Most data centres have dozens to hundreds.

You’re absolutely correct, but “quite big” might be the single biggest understatement I’ve seen in my life.

You need to think about how an infrared laser works. You’re taking electricity, converting it into light and then focusing the light.

So you’d need to take the heat from your GPUs, inefficiently convert it into electricity (a lot of it would remain as heat), then inefficiently convert electricity into light (much of the electricity would turn back into heat in this process) and then focus the light away from the space data centre.

Now, we already have a process for moving heat away from things as infrared light, without going through all those steps (which would just reduce the efficiency of the process). It’s called a radiator, and it’s how we cool things in space. That’s literally where the name comes from; they radiate heat away as infrared light. That’s why hot things glow in thermal cameras.

It is incredibly inefficient. Radiation (ie, infrared light) is, by far, the worst way of cooling things. But in space its the only option you have, because there’s no convection or conduction across vacuum.

A top end GPU puts out about 1,000 watts of waste heat. The entire International Space Station has enough cooling for 14 of those, if it was doing nothing else whatsoever. An average server rack contains 72. The ISS cost $100 billion dollars. So at a minimum you’re looking at around $500 billion to put one single server rack in space. And that’s before accounting for the heat from the sun, which we can’t avoid because we need solar power to run this thing. So probably closer to a trillion. In other words, twice the already ludicrous price tag of Sam Altman’s “Stargate” project. For a single server rack.

For anyone who doesn’t know, this is because space is an absolutely terrible place to put computers. Getting power is actually the easiest problem to solve, and is still really hard, because building any kind of infrastructure in space is hard. Then you’ve got all that radiation you have to shield against because you’re no longer protected by the Earth’s atmosphere, and worst of all you’ve got the cooling problem because Jesus fucking Christ, space is not cold!

This is why I get annoyed every time a scifi movie shows people freezing to death in space. Because it leads to this level of mass delusion and then suddenly it matters and everyone just unquestioningly believes the lie that space is cold. Space is a vacuum. A vacuum is what your Contigo travel mug uses to keep your coffee scalding hot after four hours. If vacuums are that good at keeping something hot when it naturally wants to get colder, think about what they’ll do to something that is actively generating heat. All of your components are going to cook.

There are proposals to put data centres at the bottom of the ocean that are substantially more credible than this idiocy.

Don’t buy the hype. They’re not acting in good conscience, they’ve just weighed the pros and cons and decided that the PR hit isn’t worth it.

This is just them sticking to their principles.

I’m not so sure about that one.

AI company Anthropic amends core safety principle amid growing competition in sector

AI safety leader says ‘world is in peril’ and quits to study poetry

They’re not. Conscience has nothing to do with this.

They just don’t think the PR hit is worth it.

Whenever companies choose to act in a way that we perceive as good, we were the voice of reason, not them.