Assessing China’s importance as an AI superpower

[JulianHazell]

Link post

I recently published a blog post where I tried to assess China’s importance as a global actor on the path transformative AI. This was a relatively shallow dive, but I hope it will still be able to spark an interesting conversation on this topic, and/​or inspire others to research this topic further.

The post is quite long (0ver 6,000 words), so I’ll copy and paste my bottom line takes, and (roughly) how confident I am in them after brief reflection:

• China is, as of early 2023, overhyped as an AI superpower − 60%.

• That being said, the reasons that they might emerge closer to the frontier, and the overall importance of positively shaping the development of AI, are enough to warrant a watchful eye on Chinese AI progress − 90%.

• China’s recent AI research output, as it pertains to transformative AI, is not quite as impressive as headlines might otherwise suggest − 75%.

• I suspect hardware difficulties, and structural factors that push top-tier researchers towards other countries, are two of China’s biggest hurdles in the short-to-medium term, and neither seem easily solvable − 60%.

• It seems likely to me that the US is currently much more likely to create transformative AI before China, especially under short(ish) timelines (next 5-15 years) − 70%.

• A second or third place China that lags the US and allies could still be important. Since AI progress has recently moved at a break-neck pace, being second place might only mean being a year or two behind — though I suspect this gap will increase as the technology matures − 65%.

• I might be missing some important factors, and I’m not very certain about which are the most important when thinking about this question − 95%.

[HaydnBelfield]

This was a really interesting and useful read! Posting the summary from the end of the post, as I found it helpful:

To summarise, here are the most important factors informing my view:

• China’s research on the important sub-domains of AI (such as transformer architectures and deep learning) are less impressive than headlines might otherwise indicate.

• I suspect China’s economic growth will slow down considerably unless its political and economic system changes in a more pluralistic direction (and even then, that might not be enough). This will make spending on large, speculative projects like TAI more difficult to justify politically, provided basic needs are not being met and growth is stagnant.

• China might create interesting narrow applications of AI in domains such as surveillance and other consumer products, but this might not be enough to propel their labs and firms to the frontier.

• China has a massive problem with producing compute, and the proposed solutions do not seem to be sufficient to emerge at the cutting-edge of semiconductor manufacturing.

• Some of the most promising short-to-medium-term paths towards TAI will require access to gargantuan volumes of computing power, and the US government has recently taken decisive action to prevent China from accessing it.

• First mover advantages are real: China is not in a great position to emerge at the front any time soon, and the most important actors from a governance point of view are probably the ones who are most likely to develop TAI first.

• China will struggle with talent if it primarily relies on Chinese-born scientists, who tend to return to China at low rates. A more liberal China with higher quality-of-life might attract foreign workers, but creating such a society is not exactly easy, nor is it necessarily desirable for many government elites.

Here are some things that would cause me to update in the other direction:

• China manages to avoid a huge growth slowdown, and this cumulative economic growth makes the Chinese economy truly enormous, and solidifies the CCP’s political power even further.

• China begins producing leading research in important TAI sub-domains, or is able to closely follow the West. A cutting-edge algorithmic or architectural discovery coming out of China would be particularly interesting in this respect.

• China’s centralised access to data gives them a massive advantage over the West and their pesky data protection laws in the long-run.

• Creating narrow AI products and services for the Chinese market proves to be insanely profitable, enough to flush Chinese labs with more cash than their Western counterparts.

• China solves its semiconductor struggles and begins taking a large chunk of the semiconductor market, or manages to emerge as a frontrunner in an alternative computing paradigm such as quantum computing.

• Creating TAI requires less compute than previously thought, or is possible to do with the kind of inferior-generation semiconductors that China can produce domestically.

• First-mover advantages are not as important as once thought, and the Chinese government can spend in an attempt to narrow the gap.

• China begins repatriating researchers who go abroad at much greater rates, or manages to start attracting non-Chinese talent in considerable numbers.

• The Chinese government decides to throw a much larger portion of their GDP towards AI than other comparatively sized economies, and it turns out that money can buy AI progress.

[David Krueger]

A cutting-edge algorithmic or architectural discovery coming out of China would be particularly interesting in this respect.

Kaiming He was at MSR in China when he invented ResNets in 2015. Residual connections are part of transformers, and probably the 2nd most important architectural breakthrough in modern Deep Learning.

[Nathan Young]

You should make manifold markets for all these statements and put them in the comments.

[MichaelA🔸]

Thanks—I only read this linkpost and Haydn’s comment quoting your summary, not the linked post as a whole, but this seems to me like probably useful work.

One nitpick:

It seems likely to me that the US is currently much more likely to create transformative AI before China, especially under short(ish) timelines (next 5-15 years) − 70%.

I feel like it’d be more useful/​clearer to say “It seems x% likely that the US will create transformative AI before China, and y% likely if TAI is developed in short(ish) timelines (next 5-15 years)”. Because:

• At the moment, you’re saying it’s 70% likely that the US will be “much more likely”, i.e. giving a likelihood of a qualitatively stated (hence kind-of vague) likelihood.

• And that claim itself seems to be kind-of but not exactly conditioned on short timelines worlds. Or maybe instead it’s a 70% chance of the conjunction of “the US is much more likely (not conditioning on timelines)” and “this is especially so if there are short timelines”. It’s not really clear which one.

  • And if it’s the conjunction, that seems less useful than knowing what odds you assign to each of the two claims separately.

[JulianHazell]

Yeah, fair point. When I wrote this, I roughly followed this process:

• Write article

• Summarize overall takes in bullet points

• Add some probabilities to show roughly how certain I am of those bullet points, where this process was something like “okay I’ll re-read this and see how confident I am that each bullet is true”

I think it would’ve been more informative if I wrote the bullet points with an explicit aim to add probabilities to them, rather than writing them and thinking after “ah yeah, I should more clearly express my certainty with these”.

[Robi Rahman]

A second or third place China that lags the US and allies could still be important. Since AI progress has recently moved at a break-neck pace, being second place might only mean being a year or two behind — though I suspect this gap will increase as the technology matures − 65%.

This is three different claims. Which one are you 65% confident in?

[JulianHazell]

I think I was just reading all of those claims together and trying to subjectively guess how likely I find them all to be. So to split them up, in order of each claim: 90%, 90%, 80%.