Hacker News Reader

Comment by novaRom | Hacker News Reader

I just finished reading 'AI Superpowers' and I hardly understand now if US has any single chance to compete with China's Tech fairly, openly, democratically, with respect.

View on HN

You should go back and read that book much more skeptically.

That whole book is a PR piece from a VC trying to generate deal flow. Scratch below the surface and there's not much substance to his 3 big arguments.

http://www.paulgraham.com/submarine.html

Well, thank you this was a pretty good read. I'm volunteering in a small nonprofit and do believe we work on interesting stuff but when I try to reach out to the press I feel they are totally uninterested. (Most of the time... The state/public TV usually is more open to us)

The US has overwhelming technological advantages over China, breakthroughs in AI still largely come from the US. China has more data and state support though.

...in what area is China actually leading? Catching up and leading require two very different skill sets.

Do you follow ML ? A lot of big-impact papers in Computer vision has come directly from mainland China and Hong Kong. MSRA, CUHK, Sensetime, and Baidu have made quite a few advances in the field.DMLC a group of members composed mostly of Chinese students basically wrote things like MXNET and Xgboost.

Granted most of these are collaborative ventures, but calling China 'catching up' at this stage is ignorant.

And yet nearly all SOTA research is done in the US.

Eventually folks in China realize they could easily be making $500K+/yr at FANG and be financially independent in 10 years, so they leave for the much, much greener US pastures, and stay here.

Here's an example, these are the golden standard. Which would be the US equivalent?

https://www.dji.com/gr/products/drones

Here's another -- better than Google by a wide margin:

https://www.theverge.com/2019/4/1/18290215/huawei-p30-pro-ca...

DJI is impressive - but is DJI leading because of AI? To me, that is mostly manufacturing. And this is indeed an area where "the west" is severly lacking. (Disclaimer: I am no drone expert, so this is a real question)

The verge says the camera is "...somehow even better than Google’s Night Sight". That is nice - but not a "wide margin".

>DJI is impressive - but is DJI leading because of AI?

No. But I was answering the parent comment "...in what area is China actually leading?" -- not specifically about AI. And those are just 2 examples I'm familiar with personally (as a consumer). There could be tons of others.

>The verge says the camera is "...somehow even better than Google’s Night Sight". That is nice - but not a "wide margin".

No, the Verge gushes over the camera throughout the review. You just read the subtitle ("somehow even better"), but missed their qualification of how much better it is in the review:

"I spent this past weekend comparing its camera against Google’s Pixel 3 and struggling to believe my eyes"

"The Pixel shot is, by any other standards, impressive. The P30 Pro is simply in another league, and the quality of its photo doesn’t require explanation."

"The less light there is in an image, the stronger Huawei’s P30 Pro advantage grows. In the following example, featuring an unlit bathroom where my eyes could detect shapes but no colors, the P30 Pro does the unbelievable by actually focusing and producing a very respectable image. With 30 seconds of editing in Google Photos, I can make it look like it was taken in the daytime. The Pixel attempt, by comparison, is out of focus. The Pixel, like the vast majority of mobile and even pro DSLR cameras, can’t obtain focus with so little."

1) Huawei phone has more cameras and sensors 2) Google's version is ~6months old

So yeah, some caveats there

I don't think Google couldn't put more cameras (or that the cameras make the real difference, this is an algorithmic ML matter -- expensive DSLRs with fast lenses and still can't do it as well).

Or that Huawei developed all this in the span of 6 months that Google put out theirs. They were developing it before Google released their version.

> fairly, openly, democratically, with respect.

Why would the US change its modus operandi ? :P

What makes you believe US completes fairly and openly? Don't you see news on China import embargo and capture of it's executives.

I don't think that novaRom was not saying that we compete fairly. I read novaRom's comment to say that we can't win if we play fairly.

I also read 'AI Superpowers' and I agree with the book's premise that long term China will win the 'AI war' economic competition because their society allows broad use of data and since they are more of an online Internet-connected society they have much more data to use for training deep learning systems. China is also throwing a ton of money at supporting startup AI companies.

I pains me to be too critical of my own country (the USA) but I think we are really making bad mistakes by: 1) not sufficiently supporting research at the government level. 2) turning our students into life-long debt slaves with student loans. 3) screwing up our long term economic and social health by, for example, not providing free meals to poor kids in schools (one of the very cost effective ways to get a more educated work force - kids who are hungry in school tend to not do well). 4) stop allowing Facebook/Google/etc. unfettered use of private data while at the same time not figuring out a way to get more training data to people doing basic reseaerch.

I haven't read the book, I'm adding it to my encumbered reading list (this site spins off so many excellent recommendations it's difficult to keep up).

1) Is the argument then that only the government can succeed with large scale initiatives? Pretty much everything that government is involved with becomes encumbered, if it's liked to a code base it drives up complexity for diminishing returns performance wise. I like the mission that NASA does and private initiatives are much more cost effective. 2) Schools are expensive with the rise of access to cheap money and ballooning administrators vs instructors. Should this be regulated? 3) I'm not against the "lunch" program. I'm curious why is it a schools job to feed students? The meal programs extended from lunch to breakfast and dinner. Isn't it a parents basic job to feed their kids? If this were someone with a pet would you consider them a good custodian if they can't meet their obligations? I'm not talking about someone on hard times I understand crap happens but more of an extended living scenario. 4) Stop giving information to these companies. This is akin to "I can keep a secret but the people I tell can't". Is it really private data if you share it? If you have a spending problem, stop buying crap.

What are your proposals to drive research domestically? If there were a market for it that's untapped seems like a goldmine.

Thanks for your comments. For 1) occasionally the government does a fine job. I worked on two large scale DARPA projects that were well managed and had good results. Also early funding of the Internet, grants for university research, etc. 2) then why can countries like Germany provide tuition free college, even for foreign students? To me, education is starting to look like big business. 3) because the free food for hungry disadvantaged children is cheap with a large potential payoff. 4) well. It would be better to have a long talk about this, but briefly: I alternate from being a long term FSF/ACLU supporter and decide that the benefits of giving up privacy are worth it. My latest ‘sin’ in this regard is signing up for G Suite and using it for my whole work/research workflow. I try to be mindful about privacy and decide in a case by case basis the value from sharing private data.

Correct, both the U.S. and China plays dirty tricks. In fact the whole world plays dirty tricks. The only difference is that the U.S. is expected to be the role model of the world, and will always look worse than others even when carrying out an equivalent dirty act. China, Russia, and others know this and they use this to their advantage.

You also don't see TMobile stealing robots from China or Facebook luring Baidu developers to "interviews" for interrogation.

You don't?

No, you don't. Zero examples come to mind.

Which US tech companies have literally attempted/stolen proprietary lab tech from a Chinese tech co?

Are brain drain interviews the same as being invited back to your native country for a job interview and being detained?

And that's just comparing magnitude, not even talking about frequency.

Reply on news.ycombinator.com