Learning Resources Play Money, 150 Pieces

There is a lot of clearer news about this latest breakthrough in deep learning technology. The author of DeepMind uses the term "self-reinforcement learning". Deep Mind likes their reinforcement learning (RL) approach, in particular, as mentioned in the article on "The Horde of Artificial Intelligence". DeepMind combines the deep learning layer with more classical RL methods to form an art form. AlphaGo Zero is the latest version of its Go-playing automation. Some people think that it is difficult to exceed the AlphaGo version that won the Human World Championship with Go. However, AlphaGo Zero broke not only the previous system but also the way to verify innovative ways. More specifically, this is what AlphaGo can accomplish.

"Research from zero" of AlphaGo Zero and AlphaZero is a big leap. In my opinion, it has the same influence as discovery of deep learning. Deep learning has found a general function approximator. RL's self-development has found widespread knowledge creation. DeepMind has not yet explained this, but another research team uses a dual process model to interpret it and call it "ExIt". This is my most ambitious forecast. The semantic gap between intuitive machines and rational machines will be bridged if it is not yet bridged. The concept of dual process theory (two cognitive mechanisms, one without model and one based on model) becomes a more general conceptualization of how to build a new AI. The concept of artificial intuition is no longer a state-of-the-art concept, but in 2018 it is a more universally accepted concept.

There are many introductory level resources that you can use to learn more about deep learning. There are lots of YouTube videos on deep learning, many of which are recent achievements, such as cat recognition, speech recognition, autopilot, video games, games, and auto-driven cars. NVIDIA has an excellent detailed tutorial, including detailed learning in Nutshell (2015). Michael Nielsen has a nice free online textbook, Neural Network and Deep Learning (2016). Presentation such as John Kaufhold (89 slide), "Brief History and Introduction of Deep Learning". For conferences and resources, see the DeepLearning.net wiki. See Quora 's deep learning, convolution neural network etc. Join Reddit's machine learning and deep learning community. For the location of work and investment, please refer to the list of about 1,000 AI companies in Venture Scanner. Some analysts estimate that about one-fifth of their solutions use or develop deep learning functionality.