Fortune Brainstorm Tech International: AI in China
- January 8, 2018
- 0 Comments
No tech conference would be worth its salt without a session on AI. The AI-inspired startups that pitched at the brainstorm tech were awesome. The panel discussion was also interesting, and in this post I’ll share some notes about the panel.
Panel participants: Zhifei Li of Mobvoi, Tom Rosamilia of IBM, and Wang Yi of Liulishuo aka LingoChamp – with Clay Chandler of Time Inc. moderating.
Data is the Key
All participants agreed the key for AI is data. The deeper and more meaningful the data set the better the learning of the AI applications could be. I was particularly interested in LingoChamp and its ability to help someone improve their English. In particular, the founder (Wang), discussed his own challenge: he could pass all the English tests but the cashier at McDonalds couldn’t understand his English. Lots of his friends had similar challenges. And you could pay a tutor and pay so much to learn a language, but how do you really finish.
Now picture how well you can train the AI algorithm when you have 50 million users speaking to your app nearly every day. That’s a lot of data to leverage. With new data sets, no reason this tech couldn’t apply to other languages and speech patterns that need correction. I liked a question asked: does it matter if the speaker is native Mandarin or Cantonese? It turns out that while “where you are from” dictates the kinds of mistakes you’re likely to make, the software determines mistakes only from your actual speaking performance – by listening to you. No shortcuts based on location, for example.
What does it mean for jobs?
Question was asked what this software could mean for jobs? Would there be fewer translators and language teachers because of this technology? Wang Yi’s perspective was that it was likely to add to a teaching program rather than replace it. To add the personalization of minute speech corrections, rather than to teach you the overall language from scratch. Also he correctly points out that there is an undersupply of qualified English teachers in China – 70% of schools lack an English teacher. So this kind of software really fills a gap in education.
How does it affect us?
And can we extrapolate from that example to how machine learning might affect the world in general? I guess it is hard to say for sure. In professions with a real lack of supply of skilled labor, machine learning might bring the work to our level, or do some of the most specialized parts. I recall friends worried about radiology work getting outsourced to cheaper labor markets – just like software. And other friends worried about losing legal work to low cost venues. But it looks like machine learning can augment both of those areas dramatically – and either make our radiologists and lawyers more productive and effective – or replace them in some respects – as the ticket-lawyer-bot example has shown.
Unfortunately, no one knows the ultimate way it will play out, so we’ll all just have to learn along the way, together!