I didn't read the article (it's paywalled and I don't want to have Yet Another Account Somewhere just to read it), but according to a Wikipedia article on air pollution in Delhi, the main sources of air pollution there are:

1. motor vehicles
2. wood-burning fires
3. cow-dung cake combustion
4. agricultural fires (crop burning)
5. diesel generator exhaust
6. construction site dust
7. burning garbage
8. illegal industrial activities
9. thermal power plants
10. cooling tower mist emissions
11. landfill files
12. road dust
13. concrete batching

14. ... so replacing their fossil-fuel power plants with solar would help and should be done, but I think they'll need to do quite a bit more than that to really solve their problem.

Cool, now that we've got the obligatory mean-girl style personal attack out of the way, is he wrong?

Would you claim that Musk loudly and publicly allying himself with Trump, the GOP, DOGE, and various ultra-right-wing European political groups, and regularly spouting off online like a 13-year-old 4chan edgelord, was a wise course of action, when he's also the person in charge of a company whose income stream depends on selling electric vehicles to environmentally-conscious liberals?

In my city, a large percentage of the Teslas now sport explicitly anti-Musk bumper stickers. It sure doesn't look like Tesla is going get a lot of repeat customers going forward, and Musk hasn't had much success marketing his EVs to the right-wing crowd, either. But humanoid robots, right? That's the ticket. Everyone's going to want to buy one of those, because reasons.

What will happen is that this level of quality will become uninteresting to everyone, because its ubiquity will devalue it and it will become associated with "cheap AI slop". So even if a movie company does go to the trouble of making an action movie with intense fight sequences, it won't get a lot of viewership. Why buy a movie ticket for that, when you can see infinite variations of the same thing at home for free? As Syndrome put it, "when everyone is special, no one is".

What will likely become valued instead is good characterization and engaging storytelling... at least until an AI comes out that can replicate and commoditize that too.

This is mildly interesting, but if they removed the VR teleoperation and instead required the robots to include an AI that they have trained to box (and ideally, also trash talk its opponent) it would be a lot more interesting.

It's 2035, and as you're walking down the street, a stranger furtively motions towards you from a doorway. "Hey buddy", he says. "I got here a 10TB thumb drive filled with humanity's 2 million most popular songs. It's yours for $50. Buy this and you can listen to music 24/7 for the next 22 years without ever repeating a song".

In a moment of weakness, you agree to the purchase, and it turns out the stranger was true to his word -- all the promised music is on the drive, pirated and mp3-compressed for your listening pleasure.

Now the question is, how do you go about ordering the playback? Random shuffle? Start at A and work your way to Z? Keyword search? Task an AI to choose a good mix for you? Something else?

If you consider only the climate itself, then it probably won't be utter hell -- large portions of the Earth will still be perfectly livable.

But at the same time -- large, currently highly populated portions of the Earth will no longer be livable, and all of those dispossessed people are going to have to go somewhere else, and compete for the remaining resources of the places that remain livable... which means refugee flows, and famine, and xenophobia, and violence, and war. That's where the utter hell is going to come from. Too many mouths chasing not enough grain.

I wouldn't describe the deaths of billions as "win/win". More like "lose/lose/lose/lose/lose/lose/win/lose/lose/lose/lose", maybe.

The thing is, they wanted to be deceived -- that gave them an out. Now they can go to their graves with a clear conscience, because they "know" global warming is a myth and therefore they didn't really doom their grandchildren. That's all they wanted, is some comforting lies that would give them permission to not worry about it.

You are indeed being overly cynical. You're right that they don't want to eat those costs, but you're missing that they also don't want to lose market share (and therefore sales) to a competitor who is able to charge less because the competitor doesn't incur those costs.

Which is to say, if there is an alternative way to provide the same (or similar) product cheaper by reducing/avoiding expensive CO2 emissions, they'll switch to that, as a way to remain competitive. Which is the desired outcome.

Well, now you've done it -- we have to estimate how water Dennis Ritchie used in the process of writing his C compiler.

- According the Wikipedia, Dennis Ritchie developed the first C compiler in 1972 and 1973, so I'm going to call it two years.

- The average American uses 80-100 gallons of water per day (including drinking water, toilet flushing, showers, cooking, laundry, etc).

So that puts the water usage for Ritchie's C compiler in the ballpark of 65,700 gallons total. Dunno how that compares to the AIs' usage, but there's no inherent reason why an AI needs to use water at all; closed-loop cooling systems are quite doable.

So, Musk could just add some nVidia GPUs into future Starlink modules and call it done?

I wonder if it would help at all to build one modest server into the back of each solar panel, rather than trying to concentrate all of the servers together?

One option would be to just not use orbital servers for jobs where low latency is important. E.g. if you want your server to spend the next four hours optimizing an elaborate programming solution, you can have that done in space, but if you want it to remote-control your Tesla's driving, you'd better use a ground-based server instead.

Realistically, though, latency is the least of the problems with Musk's idea. Other factors will keep this project on the ground.

Well, it's a thing, in that Teslas are driving with it now, but more specifically will it ever be safe enough to be fit for purpose (i.e. to drive actually unsupervised)? To get there, Tesla would need to come up with a way for the cameras to handle bad visibility (such as driving directly into the sun), and it's not clear to me that there is any good solution to that except adding other kinds of sensor for redundancy.

.... amortized over a 30-year operational lifespan, with near-zero risk of catastrophic loss during that period.