The US can't let China have a moment.

What's important is what you do when you get there. The US raced to the moon the first time then just quit. China, at least as far as they've announced, seems to be planning to go and actually do some stuff like set up a long term research base. The US was also planning to do some interesting long term stuff but has cancelled most of that in the interest of beating China.

I assume because they're charging by the minute they have to start up a VM or something to monitor it. They probably use the same process as any other "runner" except this one sshs to your machine and executes commands instead of doing it locally.

I don't see why someone who objects to this couldn't just replace it with their own system where their local machine does whatever it needs. Make a runner that pings your machine and then exits if you absolutely need to.

The Soviets used metric.

Do you mean the Johnny-come-lately Americans? I guess they were operating on Imperial time. They did bring a bunch of very useful 105 and 155 mm Howitzers though.

Not a special legal category that is less than, usually a lot less than, the "minimum" legally allowed.

The US: if we have to pay people enough that they can afford to live and have health care then the industry will die!

The rest of the world: ???

We build a fair number of metal things that spend their lives in sea water. Most of them have heat exchangers too.

Microsoft did that. It apparently worked okay. They've ended the experiments and haven't decided to deploy it in production.

Individually directly paid taxes

Piling on some more qualifiers? You are probably correct that the rich pay most of those as well, but if so then why the misleading generalization?

Probably because the number isn't as big. Income taxes are (usually) designed to be progressive. That's the point. Most sales and property taxes, including the hidden ones, are only weakly so. Billionaries don't eat thousands of times more bananas than poor people.

You don't need to "think", whatever you think that is, to do depth from motion processing. You don't even need it to do contextual range estimation from still images. There are perfectly good old fashioned engineered solutions for both. Learning algorithms just let you do it faster by learning some shortcuts.

Just curious, are we measuring this by sales or sphincter size here?

I would think pressure would be the appropriate metric. Since it's Ford, I guess that would be PSI.

The author pretty clearly dropped that big about "used to build the Nagasaki bomb" to spice it up a bit. I'm sure the RTG also contained some steel, which was also used to build the Nagasaki bomb.

It's not. Kodak sold a pile of 1000 odd patents to a consortium of tech companies for something like half a billion dollars. It was the accumulated bullshit of thirty years of "transferring a digital image via digital computer network on a Tuesday" type stuff, plus some actual innovation, much of it in algorithms, mixed in.

Hacky: the first of pretty much any gadget is going to be a pile of junk on some engineer's desk. So long as it fits on a desk.

Why secretive? He took an image sensor invented at Bell Labs and manufactured by Fairchild and used it to make an image. Secretive from his bosses maybe, since he managed to put Kodak out of business.

Income taxes, not taxes.

"Instant" profit is more like "in five years" or "during my time in charge" profit. That AI "research" is very much aimed at generating profits in the short term. Nobody really cared that much until OpenAI announced something that could potentially cut into Google's ad+search money pipe, then the race was on. It follows precisely the silicon valley software company strategy: write some software, offer it free or steeply discounted at a loss to get users (i.e. "scale") then monetize it with ads.