Note that this question DOES NOT include importance via physical chip manufacturing and supply chain, which is why Taiwan and The Netherlands are not listed. To avoid confusion and get at the question being asked, that is explicitly excluded.
If unclear, will resolve via best available consensus mechanism. Will pick one winner even if unclear.
Will add additional countries upon reasonable request.
Motivation: Claim in Economist that UAE might be third most important (https://twitter.com/shashj/status/1727991912567583052)
1,000
3.00@nsokolsky If all Indians return to their home country, absolutely yes.
But at least for now, it does not seem to be the case
@nsokolsky If we were talking by 2050 I would agree. But how is India going to become an AI powerhouse in a matter of five years? Current gdp/capita even at ppp is below that of Morocco or the Philippines. The talent that doesn’t leave for the US or Europe will most likely concentrate on sectors far from the cutting edge.
@mariopasquato GDP per capita doesn’t matter. What matters is the raw number of talented engineers, of which there’s no shortage in India. And thanks to the internet anyone in the world can become an AI expert now.
@nsokolsky Raw number of talented engineers matters as long as they end up employed in the AI sector on cutting edge stuff. I don’t see that happening in India on this time frame because the low gdp/capita makes it easier to add value in other sectors. Imo of course
To clarify my position: in a developed economy like that of Japan or some European countries it’s comparatively harder to add value by providing lower tech goods or services than in a developing country. India by many metrics is still a developing country. Therefore I expect them to become a key player in hi tech only if this were justified by strategic reasons. At present it’s not, unlike in the case of China. So no economic reasons and no strategic reasons means India won’t be a key player for now. An exception to this reasoning could be that opensource AI really takes off, so that small orgs or even individuals can play a key role. There is a market on that elsewhere.
Seems worth noting that the US export controls on cutting-edge AI chips were extended to also include the UAE and Saudi Arabia last month, though with a "presumption of approval", meaning exporters now need a license to send Nvidia A100/A800/H100/H800 accelerators there, but are likely to be granted such a license. https://www.federalregister.gov/documents/2023/10/25/2023-23055/implementation-of-additional-export-controls-certain-advanced-computing-items-supercomputer-and
The UAE ordered thousands of Nvidia chips last summer. https://ghostarchive.org/archive/3NFuc
Of course AI researchers in the UAE (many of whom are Chinese nationals) can still also use compute via foreign (e.g., US) cloud providers.
ETA: The license requirement comes with a presumption of approval, which I missed at first.
@erwald Can’t Nvidia come up with a scheme where they sell the cards to a European intermediary who in turn sells them to the UAE?
@mariopasquato Nvidia would not come up with such a scheme, but yes, this sort of thing happens, and could also happen here. In fact, that is how chip smuggling typically happens. It would likely involve smaller quantities and higher prices, though, if it does happen.
@AaronHahn I'd assume where the work is done matters more. If direct control is involved that might matter.
Registered that I picked Japan, partly out of explicit boosterism (I lived there for 20 years) and partly because I think the English-speaking world greatly underrated Japan in most things and then in AI specifically.
As an example, ask any AI company or project of your choice to list their top three markets by use.
@patio11 What might be a good search string that those unable to read Japanese could put into the googles to begin learning more? (Or is it just the sort of blatantly obvious strings, like "japan AI", that are best here?)
@DaveK A few of the search strings that would work disclose the contents of conversations I may not have permission to disclose.
As a quick benchmarking method, you can either go to gwern.net and read his
list of projects he considers state-of-the-art in anime-inspired creative AI then search those project names in Twitter and observe what tweets come up, or you can do a search in Japanese for words you know like LLM and look at obviously high-status answers then ask ChatGPT to translate for you. Then you might also reflect “Hey, ChatGPT seems to be… really surprisingly good… at this language. Huh how odd.”
You can also read about the long history of Japanese economic planners planting a flag in “Japan will win robotics and AI”, though they only have a reasonable claim to success on robotics.
@trixwit It's not as implausible as it sounds. Dubai is a massive magnet for international talent (and fast growing), akin to Singapore/Switzerland, which by the way should also be taken seriously. It does not take that many top-tier AI researchers to be a global leader. UAE can throw a lot of surplus oil funds into AI and attract people from Europe and Asia (the Falcon developers are mainly French and Dutch).
@trixwit The emirates got the first non-US frontier AI model according to this question: https://www.metaculus.com/questions/16640/non-us-frontier-ai-lab-before-2026/#
@Tripping This becomes ambiguous. I don't think the UK 'has' Deepmind any more, it belongs to google. It has an office in london, but probably can't be influenced very much by the UK. When they reach AGI, the US would block the UK from nationalizing it - I'd guess it'd remain fairly independent.
@LuEE If this resolved today I would expect to pick the UK but would check others' opinions to be sure.
@ZviMowshowitz Canada is probably more relevant than the UK right now imho (e.g. by number of researchers; Bengio’s group)