... when the AI bubble bursts and 90% of the AI data centers go dark. The AI that remains will be the AI that is doing something useful enough to be worth the electricity it sucks down.

What the world would really like is something that performs like nuclear fission (lots of 24/7 reliable baseload power, deployable anywhere) but without the big upfront expense or the catastrophic risks (pollution, storage, proliferation) to manage.

Is there such a thing? Could there be? Nuclear fusion might be one answer, and they've made good progress, but it's still a bit iffy and even in the best-case scenario it won't be applied at scale for some years yet. Geothermal is seeing some interesting developments that might allow it to be deployed more broadly, so that's what I'm currently geeked over. Short of that, there's always good old-fashioned renewables+lots of storage, which can be made to work, but requires a lot of infrastructure.

I wonder, are there any computers that both have a PS/2 port and also can run Windows 11?

Au contraire -- quite a few police have a well-known back door that you can exploit with nothing more than an envelope full of $100 bills.

I suspect the idea they are hinting at is that ancient yeast might have been living on (then-livable) Mars and got meteorite-blasted over to Earth, and survived the trip. For a temporary trip like that, survival is sufficient.

The article claims to measure the severity of a memory leak defect based on the amount of memory it leaked -- but most memory leaks (that are severe enough to be noticed) are small leaks that occur at regular intervals, meaning that the program's memory footprint will continually grow larger over repeated operations.

Therefore, do you want a 1MB memory leak? Run the program for a while. Do you want a 1GB memory leak? Run the program for that much longer. Keep going, and you can eventually get to any number you want, to post in your Substack article; this makes the reported numbers arbitrary and therefore meaningless.

TL;DR: Memory leaks are a problem, and they can be avoided with care and proper coding techniques, but claiming that software quality is worse now because the leaks "are larger" is silly.

OTOH there are actual Apteras on the roadways... only about three of them, but they do exist, at least in prototype form.

In general, nothing. But rich people buy sports cars for the same reason less-rich people buy video games that let them pretend to race sports cars, i.e. so they can have fun "going fast". Driving a sports car without cool engine sounds (however you want to define them) would be like playing an auto-racing game on mute -- less fun than it might otherwise have been.

It seems it can (although whether it's economical to do it at scale is another question).

AI has invented a new variant of Pokemon Go. Gather points by taking virtual photos of imaginary destinations!

If the automated coding tool is reliable, you wouldn't need to be skilled. OTOH if the coding tool keeps emitting code that contains bugs or misfeatures, then someone will need to analyze and debug the emitted code, which is a skill. In some cases, that might requires more skill than simply writing the software by hand.

It's surprisingly similar to AI in that respect. Both technologies have been shown to work in principle, but neither of them has been shown to turn an actual profit, yet.

Yes, those are all great. And geothermal is looking really promising too, with microwave drilling technology potentially enabling it in locations where geography previously made it impractical.

Who said anything about waiting? We should be (and, broadly speaking, are) deploying renewable technologies now, and simultaneously developing fusion technology for later. There's no need to do just one or the other, when we can and will do both in parallel.

... is a movie trope where everyone in the world has perished, except for the protagonist, who is now free to roam the world unmolested, help himself to any of the remaining resources available, do whatever he/she wants, etc.

The fantasy part is the idea that the catastrophe will get rid of all the people you don't care about, freeing up their resources for your own use, while sparing you and the people and resources that you do care about.

The people in this article can be blasé about AI killing humanity because at some level they think that they and what's important to them will be spared. Most likely, they think their wealth will save them. If and when they find out that they will suffer and die as well, their acceptance of the idea will evaporate quickly.

What's the end-game there, once there are no activities left that cannot be performed better by a machine? No more jobs for humans, and then everyone retires (in the optimistic scenario) or starves/riots (in the pessimistic scenario)?

Air-delivery can be faster for small items, but land-delivery is much more energy-efficient, since you don't have to support the weight of the robot and the payload for the duration of the trip.

That means that the wheeled bot can have a larger range, carry larger payloads, and needs to be recharged less often. OTOH it has to wait for stoplights, can only go 5-10 miles per hour, etc.