dataronin · HackerTrans

I mean, this is neat but as someone who actually tried to build a computer vision product, can I just say Open Images data aren't quite enough? Also, computer vision isn't quite at "human level" yet. For your own project, building a model that has 90% accuracy on the test set is awesome but for an actual product to be released into the wild, it could have serious problems (not to mention adversarial examples).

dataronin

9 anni fadiscuss

I am a Chinese born AI researcher and I've had most of my higher education in the US. The funny thing is that when I was a high schooler in China, I was not particularly good at math and basically hated anything quantitative. There are two points I want to make to counter some of the arguments people have brought up.

First, even though Chinese high schoolers are "better" at math and there are certainly many gifted kids, it does not necessarily mean that they are better at abstract thinking and will excel at higher levels. On the other hand, some of the most talented people I know did not have super impressive test scores. They are insatiably curious and think from first principles.

Second, if you look at the top researchers today, you'll find their strength comes from combining knowledge from different disciplines, e.g. computer science and neuroscience. The education system in China pretty much hones students in a single dimension, pushing them into a narrow corner. Also, the most important thing in research is not just understanding what has been done. It is thinking critically and challenging authority. Unfortunately, in my opinion, this really goes against the Chinese regime and Chinese culture to a large extent.

I wouldn't be surprised by a large number of future Chinese AI "engineers" who can quickly replicate the latest research. But I think China has a long way to go in order to create an environment that allows students to freely explore and think before a real AI breakthrough.

comments