Hacker News Reader

Comment by paxys | Hacker News Reader

It's pretty clear that "the most highly sought-after people in the world" in AI aren't fresh college grads or junior engineers.

If someone with a PhD and decades worth of research and practical experience in cutting-edge ML/AI/semiconductor tech wants to immigrate to the US, it's idiotic to put them in the same lottery as a $50K/yr "IT consultant" from Infosys. This is exactly why the US completely lost its edge in hi-tech manufacturing and is now losing ground in software/apps as well.

You want your junior engineers to have jobs and gain skills? Maybe don't say no to the best talent from across the world who can mentor them.

View on HN

It's strategic. China has amazing AI talent being developed. US is still generally a more desirable place to live. If that talent comes here, we'll maintain world leadership. If that talent stays in China, US competitiveness can completely implode.

But yes, junior talent has relatively little value in a winner-takes-all field, while top talent, disproportionately large value.

> China has amazing AI talent being developed

What? I'm no expert but in basically every field and endeavor, it seems to me like all the PRC does is copy others, but have little to no legitimate homegrown innovation, and lots of "research" being basically fake and fluff.

+ even if PRC was the global hub for AI innovation, for obvious reasons, I'm not sure it would be wise for the US to import that talent.

no -- the formal research literature in competitive settings are filled with part or whole PRC teams.. written in science English and often very thorough.. are those papers all advances? no, many are tiny variations on known methods, use the boiler-plate bragging language about State of the Art performance, and often cite mostly other Chinese authors as sources.. but honestly, the effort and erudition are, on the whole, impressive and competitive.

I would add that when it comes time to productize the research the code quality of open source AI tools for this from China is conspicuously high. In the west this part gets ignored too much.

About half of the best AI researchers I know are Chinese. I've visited universities there, and the quality of education is very high. There are a few barriers:

1) Language. Much of the stuff (including e.g. variable names and comments in code) are Chinese.

2) Communication style. Even in English, things are just organized differently.

3) Culture and national differences. The issues you mentioned are real; academic publishing in China is even more problematic than in the US. (We have our own issues, which are at least as severe; just different. Every place has issues).

4) Looking in US-style sources in China, and vice-versa, isn't likely to lead to optimal outcomes.

Those and others means things are often lost in translation, but at the core, once you learn to bridge those, the quality of the better work there is very, very good.

Try reading the author list of AI papers. Even those from western universities and labs. If you can't spot a pattern, then there's Sum Tin Wong with you.

> it's idiotic to put them in the same lottery as a $50K/yr "IT consultant" from Infosys.

well yes, but selective immigration is eugenics with extra steps.

Reply on news.ycombinator.com