The game is directly related to the study of decision-making support, and is one of the main contents of artificial intelligence (AI). The changes in playing Go are the most complex, so Go is considered the most challenging game in AI. AlphaGo was the first computer Go to beat the human Go champion, and we call the subsequent AlphaGo Zero, Alpha Zero and it 3$\alpha$. Their emergence is a great step in AI and an important milestone in the history of AI, which has also raised some questions that are worth thinking. In this paper, after describing the main characteristics of 3$\alpha$, we briefly comment on 3$\alpha$, such as its adscription of the success, whether they have intelligence, their research level and so on. Some cross-domain problems caused by 3$\alpha$ are considered, such as the symbolization of inductive logic, the new issues of distributed system, the intension of AI, how to implement intelligence, artificial brain, and so on. In addition, we also briefly introduce the conceptual structure of deep learning, the use method of tensor processing unit (TPU) in 3$\alpha$ and the main instructions of TPU.
Paper
Full text
Thoughts Raised by 3 Alpha
Semantic Scholar · Computer Science · 2019
Abstract
The game is directly related to the study of decision-making support, and is one of the main contents of artificial intelligence (AI). The changes in playing Go are the most complex, so Go is considered the most challenging game in AI. AlphaGo was the first computer Go to beat the human Go champion, and we call the subsequent AlphaGo Zero, Alpha Zero and it 3$\alpha$. Their emergence is a great step in AI and an important milestone in the history of AI, which has also raised some questions that are worth thinking. In this paper, after describing the main characteristics of 3$\alpha$, we briefly comment on 3$\alpha$, such as its adscription of the success, whether they have intelligence, their research level and so on. Some cross-domain problems caused by 3$\alpha$ are considered, such as the symbolization of inductive logic, the new issues of distributed system, the intension of AI, how to implement intelligence, artificial brain, and so on. In addition, we also briefly introduce the conceptual structure of deep learning, the use method of tensor processing unit (TPU) in 3$\alpha$ and the main instructions of TPU.