Baidu’s Andrew Ng on Deep Learning and Innovation in Silicon Valley | Hacker News Reader

Baidu’s Andrew Ng on Deep Learning and Innovation in Silicon Valley

blogs.wsj.com

85 pointsavgn9 comments

Having lived in China and experienced the Great Firewall of China firsthand, I've always wondered what role Baidu and other firms like it play in controlling the conversation and identifying dissidents. It would not surprise me if Baidu plays a large role in not only in providing search like Google but specifically tailoring those results to be inline with the content the Party would like the people to see and uses it's crawling and machine learning ability to identify dissidents.

This is of course speculative based on my own experience with the Firewall and seeing the degree to which companies located within the PRC need to cooperate, but it wouldn't surprise me if machine learning techniques are already used in such a capacity. Does anyone here that know a lot more about Baidu have any information supporting or refuting such speculation?

I know Google has already ventured into this territory with it's autocomplete word blacklist[0] and modified results for terms like torrents.

[0] https://news.ycombinator.com/item?id=8577513

> There are a lot of deep-learning startups. Unfortunately, deep learning is so hot today that there are startups that call themselves deep learning using a somewhat generous interpretation. It’s creating tons of value for users and for companies, but there’s also a lot of hype. We tend to say deep learning is loosely a simulation of the brain. That sound bite is so easy for all of us to use that it sometimes causes people to over-extrapolate to what deep learning is. The reality is it’s really very different than the brain. We barely (even) know what the human brain does.

100% agree. Needs to be said 100x more. What is one of the first things Ng taught in his ML class? These methods are tools, but that doesn't mean every problem requires a hammer. Because Deep Learning is getting press start-ups think they need to use it. And i really doubt that most of the start-ups that claim they are using it, are really using it properly.

It's at the point now that when i see a start-up claim they use deep learning i flat out don't believe them.

As someone building a deep learning platform as a startup[0], I do training among many other things and some of the trends I have seen in this space are that many aren't even familiar with what deep learning is good for, let alone why they should use it or where the hype is coming from. If you want a good explanation or 2 about deep learning, I would suggest a recent panel discussion by some people brought together by gigaom (disclaimer I was one of them)[2] as well as Andrew. What most people who try to sell you deep learning won't tell you is how impractical the algorithms can be to use and train. If you need that extra bit of accuracy, make sure you understand what you're getting yourself in to.

[0]: http://skymind.io/

[1]: http://gigaom.com/2014/09/22/heres-what-you-missed-at-gigaom...

>Large parts of China are still a developing economy. If you’re illiterate, you can’t type, so enabling users to speak to us is critical for helping them find information.

Baidu needs voice search because so many Chinese are illiterate? WTF! This is an incredibly bizarre statement.

He's suggesting there are a significant number of Mainland Chinese users who speak Mandarin good enough for machine transcription- but are illiterate? Anyone who has lived in China for a significant amount of time can tell you that's just not plausible.

The people who have poor reading skills (in ten years I have never met any completely illiterate Chinese) generally only speak their provincial dialect. Last check only 53% of Chinese can communicate in Mandarin, while only 5% are illiterate. The overlap is nearly non-existent.

I think Ng picked up a bit of Singaporean bias during his time there- it certianly seems he has spent no time on the Mainland.

wikipedia describes the situation here [0], but i think in general "literate" has a specific meaning here, that you need to know x amount of characters. Some cursory googling leads me to believe the rate for urban dwellers is 2000 characters, but for rural it's only 1500. According to a UNESCO study a few years ago, "11% of the world’s 800 million illiterate adults live in China."

I know it's comparing apples to oranges, but some of my friends in Hong Kong are perfectly fluent in Cantonese and Mandarin but really, really struggle with reading and writing.

[0] http://en.wikipedia.org/wiki/Written_Chinese#Literacy

Maybe they are attempting to include all dialects.

Hi Wogef,

Sorry to bother you (and everyone else) on this thread. I'm interested in visiting Shenzhen and would like to meet up. I tried to reach you on the email on your profile but it didn't work. Are you able to reach out on the email in my profile?

Regards,

Watch Ng at 57 minutes discuss the "cocktail party" problem and present an amazing solution: http://www.youtube.com/watch?v=UzxYlbK2c7E&t=57m0s

His course on ML made me interested in ML field and drew to coursera.

Reply on news.ycombinator.com