An anecdote is not correlation. It's not even weak evidence of a correlation.

Skinner, 1948, "'Superstition' in the Pigeon."

https://psycnet.apa.org/fullte...

I'm not sure why the Internet thinks Bubba is Bill Clinton.

https://hero.fandom.com/wiki/B...

Actually, the Epstein comma seems to be the worst.

McDonnel Douglas landed a rocket more than 20 years before SpaceX did. ATK, Pratt and Whitney, Boeing and Lockheed have all been involved in recovering boosters using parachutes.

The point of booster recovery is to make them cheaper. SpaceX and Blue Origin are the second wave of manufacturers, focused on cost.

You haven't watched any of the Apollo landings? They're on YouTube.

Those geniuses think that by turning a landlocked province into a landlocked country they will suddenly gain access to the ocean.

New Glenn is about the same capacity as Falcon Heavy, maybe a touch less, both with full reuse. But SpaceX has also demonstrated recovering Super Heavy, which is a LOT bigger. You could pretty easily slap an expendable second stage on Super Heavy and the highest launch capacity ever.

If my conversions of crazy American units are correct you could stick an entire fueled New Glenn on top of Super Heavy and launch it.

Right. NASA, coordinating Boeing, Lockheed, North American, Rockwell, etc. were the first movers. SpaceX is the "copy cat" that takes the basic idea and mass produces it for a buck each.

It doesn't seem likely. The AI companies would have you believe that more compute equals smarter but they're already hitting dimishing returns pretty hard.

That's almost beside the point though. Railway and fiber companies built out more infrastructure than was immediately useful but then a bunch went broke and the survivors concentrated on making a profit. The big AI companies are in that first phase where they're trying to build stuff faster than everyone else. Next inevitably comes the part where they try and actually make a profit. If that involves ever more intensive investment then great, we've got a new industrial revolution, just like your semiconductor example.

If you read the Reddit post there's no contradiction:

These are not people's email addresses and never were.

* no actual numbers considered

** or we could just move the planet a little further from the sun

People understand the idea of wired displays. They even undertstand the idea of wired VR headsets, the PSVR being one of them.

If you're not making a console with the headset as a peripheral though, ARM SOCs are cheap and they let you offer a lot of features the majority of people want, like wireless and the ability to watch, um, videos in bed.

A datacentre with lots of GPUs should depreciate the same way a regular old datacentre does. If they're calculating depreciation other than the way they do for regular datacentres that would be very suspicious.

In reality, models require vastly more computation to train than they do to use, and more still to develop, so the current spending is more accurately compared to something like the costs to construct railways, which is much greater than the costs to run them, and the asset is not the GPUs but the trained models.

Sounds like they're working on doing exactly that. Chrome, YouTube, Android, and a bunch of other Google stuff seems to use FFMPEG, not to mention any other open source projects they do this to. In 90 days Google might just be saying "hewooo Internet! Here's a vulnerability in most of our software that we didn't even try to fix!"