jjk166

> The real question is whether it would happen in a practical timeframe.

How are we defining a practical timeframe? How many billions of man-years did it take for a human to come up with velcro? In a world without velcro I could easily imagine an llm proposing the idea long before a pure human team.

Europe is putting economic restrictions on Starlink; the technical restrictions the US placed on Starshield are unrelated.

So? If you import $100 billion dollars of components and you do nothing to them before reselling abroad, you have $100 billion of exports.

> "It would cost more than one quadrillion dollars per ounce to produce gold by this experiment."

"This experiment" being a radically different method than the one under discussion. Wikipedia pulled something useless from someplace irrelevant, which is exactly what should be described as "pulled out of their butt."

The LHC is also neither a nuclear reactor, nor a gold making machine. You can make gold with it the way you can use a space shuttle to deliver newspapers. That it is possible is an interesting tidbit of trivia which you might include in an encyclopedia article, it is not however evidence supporting the claim that newspaper deliveries cost millions of dollars.

> So, in short: they want AI to take the good jobs.

You're assuming that high paying jobs are good jobs. While high pay is obviously desirable, the jobs themselves often suck, and people are paid those high premiums in part to put up with the sucky conditions. If lots of people both enjoyed and were able to perform a job, the pay would be low.

People are saying "AI took our jobs" or are concerned it imminently will. A lot of white collar tasks are doable by AI today.

Ah but I'd take roofing in the summer over septic tank cleaning any day. The existence of even shittier jobs does not make any particular job not-shitty.

Which would you consider more premium, the hotel with a lobby with marble floors or with linoleum? That doesn't mean marble provides any utility that the linoleum does not.

> remove the need for a waiter to come by and ask, or for you to get up and order, but by automating that away you also remove that little bit of social interaction for both the people ordering and the waiters.

Okay, but were you coming to the restaurant for social interaction with the waiters? Have you stopped going to those restaurants that serve you good food at a reasonable price and which are otherwise desirable places to eat because of the lack of social interaction with waiters?

People dislike places that charge too much for mediocre service, and those places disproportionately use the QR codes for ordering and don't pass on the savings from labor reduction to the consumer. Very few people actually miss their waiter interrupting their conversation to tell them the specials.

You wouldn't enrich the 198 isotope, you'd just bombard natural mercury and chemically separate out the gold afterwards.

> "...the production cost is estimated to be a trillion times the market price of gold."

Wikipedia just pulled this one out of its butt. The value of a neutron in a nuclear reactor is the opportunity cost of it not hitting a uranium atom to produce fission. 1 mole of neutrons can produce about $100,000 worth of wholesale electricity assuming 40% efficiency. 1 mole of neutrons could alternatively produce 1 mole of gold, which is worth about $26,500. So it's roughly 4 times the market price of gold.

Of course there would be the cost of building and running the reactor and the associated chemical processes to collect the gold, but these cost are probably similar orders of magnitude to mines and refineries. Certainly not 12 orders of magnitude more.

While it's not economical to make gold this way, it is economical to make certain medical isotopes this way.

Anthropic's quarterly revenue is $16 billion. Assuming they are 40% of the ai market, pretend that all the ai market is american, and pretend all tokens are exclusively sold abroad, that would make AI tokens 7.3% of US exports, comparable to aircraft (7.5%). Of course in reality only a small portion of these tokens are american exports.

A single airliner can be half a billion dollars. That's about 1/4000th of the US's total goods exported in a single item.

You wouldn't need many ships. The volume of a single shipping container is enough for 1300 tons of gold. Of course gold is extremely heavy, you'd need a specially built container to support that much gold, but the point is the storage space of any ocean going ship can comfortably carry the entire gold reserve of an entire country.

If you actually loaded gold into shipping containers and obeyed the weight restrictions, all the gold in the world would take up less than 8400 TEU, which would be considered a mid size cargo ship.

I watch a guy solve a sudoku every day. The puzzle is always different, but conceptually it's just a minor variant of the same video. I don't watch Mr Beast or that sort of content but I can't judge anyone who does.

I mean if you are of the opinion that people stop watching videos because they don't like them and continue watching because they do, then it would logically follow that by tweaking videos in such a way that people keep watching, you are giving them what they want (even if it's not necessarily what they may say they want). You don't need to be a narcissist grifter to want to make what people want. It's not like this is high art, it's mass consumer entertainment.