So 1) That's not what you said. 2) In fact, even if you had said that, it would still be wrong. The vast majority of of these benchmarks are benchmarks made by independent organizations where you can find the details of how they work without too much effort. For example, GSM-8k is available here. https://paperswithcode.com/dataset/gsm8k.

Also if one clicks through to the article, they list a whole bunch of other explicit benchmarks including some where they performed less well than the other AIs.

The summary explicitly gives examples. From the summary tests included " GSM-8k for mathematical reasoning and MMLU for expert-level knowledge." Not reading the linked article is one thing, but at least read the summary please.

No, it really is corporate interests. And it isn't a market failure as much as it is companies lobbying state governments to do things which wreck the free markets. For example, some US states have laws which make it difficult to build transmission lines into that state if one doesn't have generation in that state. That has made it very difficult to make new transmission lines for wind power from the Midwest to states in the Southeast. See for example this Wall Street Journal piece from a few years ago https://www.wsj.com/articles/building-the-wind-turbines-was-easy-the-hard-part-was-plugging-them-in-11561176010. The bottom line is that a lot of this is not market failure at all, but corporations deliberately using government to restrict competition from entering the markets. There are of course other issues, like how the US gives NIMBYs massive power to block things. But these are largely also things where the problem is not market failure except in so far as the governments, federal, state and local are stepping in already to distort the markets.

Sorry: the other article URL didn't end up in the comment for some reason. Should have looked at the preview more carefully. The other article I was quoting from is here https://www.usatoday.com/in-de... .

The paper is here https://www.pnas.org/doi/10.1073/pnas.2313661121. The West states in question are the Rocky Mountain states, which doesn't include California. Does that cause you to reevaluate?

By West in the paper, they explicitly mean the Rocky Mountain states https://www.pnas.org/doi/10.1073/pnas.2313661121, which doesn't include those three states.

The article estimated deaths by region, and found more excess deaths in the West and South of the US. Those areas are areas which by and large had lower vaccination rates. This is a pretty big problem with your claim.

From summary:

There were also people who just refused to acknowledge loved ones had died of covid, and actively tried to block death certificates from listing covid. This is discussed in for example this article . In many cases, the official records simply used whatever family members said was the cause. Some were even more extreme. From that article:

While part of this is due to lack of resources, and both TFA and the above linked piece discuss some of that, part of this is political in nature, with one end of the US right-wing deciding that covid wasn't a major issue and thus downplaying covid deaths. Unfortunately, downplaying a disease for political reasons doesn't make it less deadly.

Letting kids have a day off is not a bad idea. But another major part of this is just how hard it is to pivot. I'm a school teacher. Pivoting to a remote lesson with 24 hours notice is a massive amount of work. Give a teacher a three or four days to prep for a remote day and we'll slide it in naturally. Tell someone less than 24 hours before that the next day is going to be remote, and it is unlikely to go well.

It is true that we're unlikely to not stay below 2 C. But at the same time that doesn't mean we shouldn't stop trying to reduce CO2. Every little bit of CO2 we produce makes the long term situation worse. We cannot look at some number and say that because we are going past that number that we should give up.

Shared functionality is a really good point that I hadn't thought about. Good call.

Old style diners all over the US and even in a few other countries looked pretty similar. It doesn't take any special shared internet to have broad trends in style and organization. There are a lot of other things that cause this. If something has a lot of copies (say similar round diner seats) they will cost less to make, so there will be an incentive to buy the same seats. And people are often more comfortable with things that resemble what they are used, so the more same looking ones end up more economically successful at the margin. Put all this together and it shouldn't be surprising if the same thing happens for coffee shops today. This doesn't require anything involving the Internet. It is possible that anon's idea of what happened here also had some influence, but it isn't needed.

The title is a bit amusing, because it sounded like it was about all "science" journals rather than all Science journals until one read it. I'm reminded of how a few years ago there was a spammy vanity level journal called "Science and Nature" and someone pointed out that someone could publish it and then say "I've been published in Science and Nature."

But many many people in California want to have more housing and want to live there. The problem is that the current setup makes it possible for a tiny part of the population to functionally block everyone else from building things. And there's a vast gulf in density. Again, the supposedly dense areas have a density that is a quarter of that of Manhattan. Even if you don't want to live in Manhattan, there's a massive amount of room there.