Big US companies who have troves of user data usually have pretty strict policies about not just sticking it all through arbitrary ML pipelines for user privacy reasons.

Small US companies don't have access to enough data.

Companies in China have both lots of user data (1B+ internet users), and no privacy concerns about running ML experiments on that data (and observing user behaviour till they find success).

Hence, China will win the next wave of ML-powered social apps.

Yeah, I have seen firsthand how paranoid big US companies are about doing anything with personal data specially media which is a good thing for users. This really slows down even simple analysis where you were looking at even a Yes/No output and makes some features impossible/hard.

But they are competing with companies now which don't have the same ideological issues.

Free trade doesn't work between countries with different value systems, the values will be exploited as a weakness. Same applies with labor and environmental standards.

This is not even remotely true, I once worked at a team in a company which aggresively run user segregation and send marketing emails based on it, then they collect click data to curate more effective emails

IMO, it's likely that Bytedance is just as exceptional in its effective use of AI in China as it is in the US. IOW, ask why Bytedance is special, not why "Chinese startups" are special.

edit: and Newsdog, from the end of the article. My opinion is looking shaky.

Elon musk recently said China rocks and the US has become more entitled and full of complacement.

I kind a have to agree, even my company starting to shift a lot of development work to team in china. Their output and hard work has been amazing.

I predict Chinese tech companies will offer wages equal to American tech companies for foreign workers in under a decade. Tech companies are really going to start feeling pressure in the coming years, and I know headlines will be full of "Nooooooo China can't possibly do this! Their economy is just a bubble!" type stuff. Just like we've been seeing since the early 2000s regarding manufacturing and the alleged fake cities and housing that supposedly "propped up" the economy.

America has gotten way too comfy with the idea that it's number 1 and forgotten that it's not an intrinsic role given to it, but something it had to work for. It's starting to look a little rough.

Very minor correction: you probably mean "complacency" not "complacement". And I do agree to an extent; from where I am standing the American software industry feels a bit bloated/stagnant/wasteful/some word I can't think of that covers all three.

This reminds me a recent story in china: senior ms employees complaint against newly hired Huawei employees, who voluntarily worked overtime, and even showed off this to other senior ms employees.

Can't be that hard to do hard work when no one gives a shit about the ethics of it.

If I had to take a guess, Asia has always focused more on math in education than the US. Given that AI/machine learning is a lot of math it makes sense (they would have a larger talent pool for building the algorithms), but that's just one guess...

A lot of those are employed by US based companies and US still can outpay companies in RoW. The only thing limiting is access to H1B visas, so I doubt talent pool is the cause here.

But I would argue that designing an algorithm that appeals to the human psyche in the way that Bytedance as done requires more than math skills. Designing the TikTok "For You Page" is not a pure ML problem, its a people problem that is being solved with ML.

Personally I felt the appeal of TikTok is that the design is ML-centric - clean buttons, very clear signal, not a bunch of mixed feedback signals like in other apps. Everyone on board seems to understand the importance of the underlying ranking algorithm

It's not just AI, every change is an opportunity China can grasp. They got enough men, money, and are motivated to compete in the market. If nothing is changing, it would be very hard to challenge current monopolies, but we're living in a world that is rapidly changing.

An example is the electric vehicle. Though EV cannot replace traditional vehicles at a full scale right now, the traditional vehicle industry is not very efficient, things are not standardized or intentionally walled by monopolies. As for EV, conceptually it should be simpler, more straightforward, and more standardized, then companies can easily prototyping and mass-producing EV on a really large scale - that's what China good at, what they're not good at is they don't have recognizable branding, but there are not too many EV brandings as traditional vehicle brandings like Daimler or VW. Both government and companies see this as a great opportunity to lead these market so it's been greatly promoted.

The semiconductor industry is another field that China may want to compete in the future. Traditionally, Intel is pretty much non-challengable. However, since mobile become dominant, and Apple started to get rid of X86 and move to Arm, there might be a big change in the future. If operating systems and software are willing to support Arm, it would create yet another chance for China to participate in the competition.

Which other chinese startups with machine-learning as its core? Do you mind listing some examples?

I think he might be referring to companies like SenseTime which were in fields like image/video recognition. I think for fields like this there is a cultural barrier which prevents companies like these from being successful in US.

Say, if FB starts using its AI chops to build smart video analysis tools which can be used by law enforcement, you would probably have employee mutiny at your hands.

I was seeing a Vice video on the surveillance technologies the companies are building in China where the engineer building it was so excited to show the technology he had built not thinking about how invasively it tracks everything, keeps history etc. I would just call it a cultural gap.

You also have face recognition used in law enforcement in Britain too, though. And from top of my head, Britain was the first country to mass roll out CCTV system as well.

https://www.nytimes.com/2019/09/04/business/facial-recogniti...

But that hasn't been replicated in most Western nations. This went to court and you had a lot of scrutiny on this.

In PRC, because of cultural differences I think most people accept the crime safety net you get over personal freedom restrictions and you can't really meaningfully challenge any such move even if you wanted to. Anyway, it is the same in Singapore and people seem happy with it despite being exposed to all the Western culture too, so I think it is a cultural thing.

The thing that British were the first to roll out this but we haven't seen any advanced analysis stuff out in public after it kind of shows the apprehensiveness around it.

It's true that the lack of awareness of privacy is pretty prevalent even in industry. Hence the ease of obtaining large amount of highly curated data. But it is also true that these details are used as material for sensationalist journalism. Sometimes I can tell clearly (by understanding what they speak) the interviewees were so innocently excited without knowing how they are cut into videos. But I could be biased too.

Yeah, I don't doubt it. It could very well have been edited to look like that, either for views or sensationalism. Here is the video if you want to see: https://youtu.be/CLo3e1Pak-Y

ByteDance and iFlytek have stood out to me. I saw iFlytek's universal translator at the airport one time and it was really impressive. There's also a number of examples in Kai-Fu Lee's book "AI Superpowers" - I guess that's the book that put this thought in my head in the first place

Labor costs?

ByteDance, the company behind TikTok, pays more for AI developers than most western companies: https://crm.org/articles/3-million-pay-packages-how-a-chines....

Speaking entirely anecdotally, I suspect having a bunch of people working on an ML algorithm actually does more harm than good. It's more important that everyone in the product stack is on the same page about the point.

The Facebook Newsfeed UI, for example, has to balance so many different signals. Compare that with TikTok's UI, which is very efficient at collecting a clear reinforcement signal and only has to deal with one homogenous kind of data. It's like the entire TikTok team, not just the data scientists, focused in on doing ML really well.

I'm really impressed and kind of scared whenever I use TikTok. It really does feel like a whole different level of ML-powered addictiveness, designed from the ground up rather than having the ML part added on.