why do you think one company will determine whether the us beats china in ai or not ? Like 75% of the authors i read on AI papers are Chinese, that should be far more alarming if you really are afraid of china getting ahead.

Research from PRC (across all of science, not specific to AI) has a terrible reputation. They are rewarded for sheer quantity. You can easily find many articles discussing this phenomenon.

So the volume of Chinese AI papers says little to nothing about their advancements in the field.

Hmmm, that's the same reputation er... western science has as well.

It's really not.

It really is. Academia prioritizes quantity over quality. Western less so than Chinese, but nevertheless it's a problem in academia. Peter Higgs (the guy who predicted Higgs Boson) recently talked about it: "Peter Higgs: I wouldn't be productive enough for today's academic system" https://www.theguardian.com/science/2013/dec/06/peter-higgs-...

I regularly read really good papers that come out of China. For instance, there's great CV work out of China.

The two statements are compatible. It can be true that there is a core of really high quality research coming out of China, and also that there is a huge “long tail” of low-quality research that is probably motivated by badly-calibrated publication quantity metrics. The US seems to be slightly more limited in this because (possibly) we have a smaller population of researchers, but also because our government funding sources for research (places like the NSF) tend to have some peer review that cuts against this sort of metric.

Huh, that's exactly what I heard about western institutions as well.

That's a problem in all of science, and Chinese research is quite good in measures like citations as well, not just quantity of papers.

Chinese papers are (with much higher probability) citing Chinese sources. It's a self-empowering cycle, which doesn't say anything about the quality.

Yes, and American papers are much more likely to cite American papers. Science is more international than the vast majority of professions, but there are absolutely still state cultures that are just more likely to have read research in their language, published by someone who's a friend or a friend of a friend, or have national institutions which concentrate scientific talent that make scientists be colleagues. Nowhere near as strong of an effect as other jobs, but it's still there.

Ethnocentrism is ethnocentric.

It's like how historical American medical data collected by universities has been misapplied to pharmaceutical and medical practice because of demographic bias. Research participants largely matched the demographics of the university: healthy white males.

Or more broadly, whenever you see a "last name" requirement on a form, you know it's software made by people who think it's normal for people to have "last names", and that everyone should know what that means.

This just in:

Researchers are vastly more likely to read, and therefore cite, papers in languages that they understand fluently.

Virtually all of this work is published in English though, even from the Chinese researchers.

You’re taking Hinton at his word. Maybe he was forced out of Google for doing nothing with LLM tech for half a decade.

I don't know if it's bad or good for the long-term interests of the humankind, but right now it feels like a Klaus Fuchs moment.