Google AutoML
Well, I’m enrolled in the class ABE 598 Autonomous Decision Making. But I’m not sure if I’m going to take it. Instead I will audit and learn what I can appreciate and share it here. Prof. Girish Chowdhary has critical attitudes to many things, especially the hottest tech news. He wants us look at things in his way, being kind of cynic. One idea that inspired me is, quoted from Dr. Chowdhary, “when you see lots of news boasting the idea of deep learning, go and check where they come from, and how do they live on. Deep learning is what medias like Google and Facebook now live on. And then think about why they want to sell you these fancy ideas”.
Yesterday, Google announced a new product called AutoML, saying to free people from tedious machine learning training task, and make everyone use the technology much easier. However, like what Dr. Chowdhary has said in class, data labelling, which is one of the most important prerequisites of ML tasks, is totally neglected in their announcement. The most time-consuming labor work are still left there while the algorithms behind is not magic anymore. The incentives of launching this so-called ML tool is unclear. What’s dirtiest that could be thought of is that Google wants free data, free labelling, and free feedbacks from the public. The intuition behind the project, as I could guess, is probably the fact that if the smartest deep learning experts are just blindly tuning hyper-parameters without knowing why, then why not just let computers do this. Apparently, the fundamental machine learning and deep learning theories are a bit behind their applications, and how deep learning works is still in a black box. However, it doesn’t mean that it’s the best way to give it over to machines and ask them to try, try and try until you get what satisfy you. Yes, it’s reasonable for a company who only wants to make money and investors happy that if only it works, it’s done. But isn’t it scary that you know nothing about the thing you create that knows a lot about you?
In addition, the stock prices of Google, Facebook, Amazon and a number of other tech companies are incredibly increasing when there are no better products or services emerging. It’s not reasonable to me at least. What I don’t want to guess is that the majority of us, including you and me, are totally controlled by the medias, which live on some fancy ideas, and want to sell the ideas to you because of the pressure from investors. It reminds me the tulip bubble… Is the same thing coming? If so, we’d better prepare for it.
News are not news anymore, most are ads and lies. Information gaining is probably is driver of economic growth in this era. I should say it’s harder and harder to tell the truth from the fantasy, the good from the bad. But I believe the day will come, when everything is on its right track with its value not overestimated or underestimated. And we, human, become more rational of knowing what we can do and what we can’t.