Hacker News Reader

Comment by klrr | Hacker News Reader

I think China will create an AI bubble, currently studying in Beijing and all EE and CS students I know focus all their time at either stats, ML or data mining. While I think China might have a big shot at the AI race, I don't think it's useful to have to only focus on AI because it's hot right now, they may lack progress in other fundamental fields that may lead to more important breakthrough and technological change.

View on HN

It's not just China. At my German university all courses related to AI/ML/data mining/stats are oversubscribed by a factor of 2 to 5. One professor went so far as to stress that he didn't have any experience with deep learning and he'd only cover Bayesian statistics in order to discourage students. I don't think it worked.

I'm pretty sure that there's a similar surge of interest in all countries.

>>I'm pretty sure that there's a similar surge of interest in all countries.

Yup. Same here in India.

People talk as though singularity is next year or so, and like Y2K jobs they have to do AI programming jobs.

Went to campus hiring like a few weeks back. And everyone had one or two show-and-tell AI/ML projects on their resume. Like every one. Everything from fitness apps to face emotion recognition.

I could be quite wrong but wouldn’t a course in Bayesian statistics be a lot more useful for someone interested in inference or however you will say it?

It would certainly be useful. I'm interviewing a lot of people for ML roles, and I've noticed a new group of "deep learning only" people coming out of colleges. They know their DL, but have no exposure to what I guess we're calling "traditional ML". Since DL isn't appropriate for all problems, they don't have the skills we want.

Those students will be awfully disappointed when they graduate and discover that most commercial software development work involves just shuffling bits around with no AI in sight.

We’ve been here before with Japan in the 80s, and it led to an extended 2nd AI winter.

China’s efforts are much larger than Japan’s were, and they have a bigger housing bubble to go along with that to boot. It is interesting to see how history repeats itself.

AI actually makes money now. There will be no more AI winters as it can now support its own development without patronage.

AI made money before.

All it takes for an AI winter to happen is for the hype to go way beyond the reality, the AI doesn’t have to not make money, it just has to make not enough money as expected. It is very much like a housing crash: the houses are still useful to live in, just that their value is not as high as expected.

AI bubble, currently studying in Beijing and all EE and CS students I know focus all their time at either stats, ML or data mining

Yes everyone wants to mine the data, but noone wants to generate or gather new original data in the first place.

I think China has plenty of data as well, of its own people at least. It's a surveillance state, and the big corps work closely with the government and so also the universities. My point is more that they have a chance of winning the ML race, but that doesn't matter if there will be more important technological breakthroughs in other areas that will have greater impact on AI than ML.

Comment of the year

China does want to gather data, it has deployed tons of sensors and cameras to monitor its population's habits on the street. With the personal credit score implemented it can monitor its population even more thoroughly. How much data it gathers on its population thanks to the Great Firewall we can only speculate.

Granted, population data is one thing since you can incentivize people to share it. Data gathering that requires investment of money into sensors and time for careful placement of them is a completely different thing entirely

> they may lack progress in other fundamental fields

Well if they crack general intelligence maybe they can generate, copy and paste researchers in those other fundamental fields.

Lack of substantial practical applications might make it a substantial waste of time and money.

Does anybody remember the 5th gen computing stuff from 30 years ago?

AI is general use. This is a good bet.

> all EE and CS students I know focus all their time at either stats, ML or data mining.

I quickly checked zhihu.com's hot topics related to CS, zhihu.com can be regarded as the Chinese counterpart of quora.com

the topic AI has 1.34m followers https://www.zhihu.com/topic/19551275/hot

the topic e-commerce has 1.14m followers https://www.zhihu.com/topic/19550780/hot

the topic database has 0.8m followers https://www.zhihu.com/topic/19552067/hot

Reply on news.ycombinator.com