Recently, AlphaGo Zero taught itself to go with unsupervised learning and beat all previous versions of the game 100-to-0 in 40 days. But the 100:0 record seems to confirm AlphaGo Zero’s unlimited potential and great contribution to the overall development of AI technology, but in fact these claims are exaggerated.

Is AlphaGo Zero’s technology scalable?

The answer is none or very limited.

AlphaGo mainly deals with man-machine combat through supervised learning and reinforcement learning. In simple terms, AlphaGo does a lot of exam questions in a short period of time through technical means, and then wins the game with clear rules. The new version of AlphaGo Zero is different from the previous version in that it learns from itself and achieves accurate “closing” without “real” questions, thus achieving victory.

In fact, a lot of people during the Period of AlphaGo, will think of the question is why go? First of all, Go is the most complicated board game with clear rules, so AlphaGo, as a representative of artificial intelligence, can win the impression of high intelligence to a large extent by defeating human representatives. On the other hand, ai has a greater chance of making successful predictions in Go because there is no data noise. You know, in image recognition, three out of ten thousand noises can make a big difference, like identifying a bear as a gorilla.

So could AlphaGo’s technology be applied to other scenarios? The answer is no.

Could AlphaGo be The next Google project to shut down?

A few years ago, when smart hardware, VR and AR were very hot, the argument that wearable devices would replace mobile phones was widely heard, among which Google was an important push. Google Glass can use eye movements and voice commands to make phone calls, take photos and other functions. Compared with smart phones, Google Glass does not have much innovation, and it is difficult to bring users a better experience in human-computer interaction. In addition, Google stopped the project in January 2015 due to problems such as power consumption, heat consumption and high price.

Google shut down the program also not only this one, at the time of 08, Google explored by users search for large data to predict flu, published in Nature, the thesis points out that based on users’ search behavior can predict the overall trend of flu, moreover through the contrast with the United States centers for disease control and prevention of influenza surveillance information, can achieve a more accurate prediction. The actual deviation was a disappointing 140%, almost completely wrong. Soon after, the project, which had been called a model of big data flaws, was also shut down.

At present, it is difficult for us to judge the actual value and future prospects of AlphaGo and AlphaGo Zero by winning or losing a match. But we need to acknowledge that Google has been trying to do things that no one else has tried to do, and its role in moving the entire AI industry forward cannot be ignored.

I believe that in addition to AlphaGo ZERO’s amazing performance, The ability of Chinese AI companies to integrate technological innovation into demand scenarios will also receive the same attention.