How does the Alpha Dog evolve?

It is necessary to combine Alpha GO with medical treatment, robotics, etc.

AlphaGo is a Go AI program developed by David Silver, Aija Huang and Damith Hassabis of Google's DeepMind company in London, England, and their team.

In October 2015, Alpha Go defeated Fan Lu, a European Go champion and professional second-dan player, 5:0. In March 2016, he challenged Lee Sedol, the world champion of Go and a professional nine-dan player.

According to the schedule, the five games will be held on March 9, 10, 12, 13 and 15, and even if one side wins three first, the full five games will be played. In the end, the "war" ended 4:1.

On March 27, 2016, AlphaGo confirmed to challenge StarCraft 2. Since the evening of December 29, 2016, an "online chess player" registered as "master" and marked as South Korean Jiudan has successively "kicked the hall" Yicheng.com and Yehu.com. From the evening of December 29, 2016 to the evening of January 4, 2017, the record of the master against the top human masters was 60 wins and 0 losses.

Before the final game, the master revealed that "he" was AlphaGo.

From May 23rd to 27th, 2017, Ke Jie and Alphago played against each other at the "China Wuzhen Go Summit". On May 23rd, 25th and 27th, Alpha Dog defeated Ke Jie three times. After the game on the 27th, it was announced that Alphago would retire.

On October 18, 2017, the DeepMind team announced the strongest version of AlphaGo, codenamed AlphaGo Zero. Its unique secret is "self-taught". Moreover, it is from a blank sheet of paper, learning from scratch, and becoming a top master in just 3 days.

The key to Alpha Dog's ability to learn successfully without relying on human knowledge is that Go has definite rules and is an objective game. There is no need for subjective judgment by humans, the machine plays according to the rules of chess, and the final game has a definite result.

In this way, Alpha Dog's learning does not require human intervention, and it can be completely automated to carry out a large number of practices. The success of Alpha Dog is a breakthrough in the self-learning method, and it is also the success of the practice test of philosophy.

Human social activities or AI games need to be practiced to continuously improve efficiency and performance. Practice always needs to be carried out within certain rules, which is the foundation, such as a stable society, the modern banking system, the game of Go, and the rules of the final game. When it comes to practice, human beings instinctively draw on some of the experiences of their predecessors as a way of thinking.

Unlike previous versions of Alphago, this version of AlphagoZero is completely unfettered by the human mind, and is therefore considered an innovative artificial intelligence. Alphagozero's capabilities have been qualitatively improved. The biggest difference is the trace, which no longer requires human data.

In other words, it has never been exposed to human chess games in the first place. The R&D team just let it play chess on the board at will, and then play its own game.

Function: 1. AlphaGo itself is an ATIFIALINTELLIGENCE developed by Google, that is, AI, which calculates the approximate level of the disconnected person according to a certain method, and then makes the corresponding level of operation, including consciousness. In this way, Cheng Yingzhou can be generous enough to solve the phenomenon that some people are offline, and the existing AI pits are all over the field.

2. Its unique secret is "self-taught". Moreover, it is from a ridge of vertical blank paper, zero foundation learning, in just 3 days, become a top master.

Alphagozero uses a new reinforcement learning method to turn itself into a teacher. At first, the system didn't even know what Go was, but started with a single neural network and played a game of itself through the neural network's powerful search algorithm.

With the increase of self-game, the neural network gradually adjusts to improve the ability to take the next step and finally win the game. What's more, as the training deepened, the DeepMind team found that AlphaGozero also independently discovered the rules of the game and came up with new strategies, bringing new insights to the ancient game of Go.

After 3 days of self-study, I beat the old version of alphago

AlphaGo Zero also differs significantly from its predecessor in three ways.

Alphago-Zero's training timeline.

First of all, AlphaGo Zero only uses black and white on the board as inputs, while the predecessor includes a small number of artificially designed feature inputs.

Second, alphagozero uses only a single neural network. In previous versions, alphago used"Policy Networking"to choose the next move, as well as to use"Value Network"Come to the winner after each move. In the new version, the two neural networks have been combined into one, allowing it to be trained and evaluated more efficiently.

Third, alphagozero doesn't use a fast, random walk-through method. In previous versions, AlphaGo used a quick move method to determine which player would win the match from the current situation. Instead, the new version relies on its high-quality neural network to assess the situation in chess.

Alphago several versions of the ranking situation.

According to Hassabis and Silva, these differences have helped the new version of AlphaGo improve the system, and the algorithm changes have made the system stronger and more effective.

After just 3 days of self-training, AlphaGo Zero defeated the old version of Alphago, which had previously defeated Lee Sedol, with a record of 100:0. After 40 days of self-training, AlphaGo Zero beat the AlphaGo Master version again.

master"He has defeated the world's top Go players, including Ke Jie, who is ranked No. 1 in the world.

For Deepmind, whose mission is to use artificial intelligence to promote the progress of human society, Go is not the ultimate meaning of AlphaGo, and their goal is always to use AlphaGo to create a universal and ultimate tool for exploring the universe. The improvement of alphagozero has allowed Deepmind to see a breakthrough in using artificial intelligence technology to change the fate of mankind. They are currently actively working with healthcare facilities and the power and energy sector in the UK to improve patient care and energy efficiency.

The pace of the times is getting faster and faster, and the speed of scientific and technological progress is getting faster and faster......

Perhaps one day, the R&D of technology will be replaced by an organizer and a group of intelligent robots ......

In this era, technology is making life more convenient.

What does the Alpha Dog portend in an evolution?

This has to make people sigh that with the progress of science and technology, in the face of the rapid development of artificial intelligence, what will your home look like? What exactly is a smart home? What does it bring us?

Smart home is the use of network communication technology, automatic control technology, audio control and other technologies to create a higher security, comfort, artistry, convenience of the residential platform, to achieve more efficient family schedule management. With the advancement of science and technology, smart home has also innovated from the previous lighting remote control, electrical remote control and electric curtain control to home security alarm, background**, video intercom, access control fingerprint control and other fields. It is not difficult to see that smart home covers almost all traditional weak current industries, and it has attracted various brand companies to enter the smart home market.

For lazy patients with advanced cancer, cooking is undoubtedly the biggest challenge, so it would be the happiest thing to have a smart hand to help you make a delicious meal. The UK's Moley Robotics company showcased its latest black technology - the robot kitchen system, which claims to be "the world's first automated kitchen" to achieve a fully automated cooking experience, and the little things like cooking will be left to you in the future. Not only can he make 2,000 kinds of food, but he will gradually increase the menu.

You can even ask a real master chef to cook, and the robot's vision system will capture and record the master's movements, time, and then generate a set of your own cooking routines, and finally clean up the kitchen, all of which can be controlled by the robot at your fingertips. It's against the sky! There is wood, there is ah!

That all changed when people were still raving Alpha Dog in May this year, and Alphago was once again featured in the world's top scientific magazine, Nature. It turned out that on October 18, local time in London, the Deepmind team announced the strongest version of AlphaGo, codenamed AlphaGo Zero.

It is said that this alphago zero defeated the original alphago by an overwhelming 100:0 advantage, and unlike the original one, this time its unique secret is "self-taught". Learn completely by yourself, learn from scratch, and become a top master in just 3 days.

The previous version of alphago was self-trained by combining the scores of millions of human Go experts with supervised learning for reinforcement learning. After months of training, relying on multiple machines and 48 TPUs, AlphaGo Zero's capabilities have been qualitatively improved on this basis, it no longer needs human data, and only uses one machine and 4 TPUs, which greatly saves resources.

Zero's success is mainly due to the reorganization of the updated neural network and search algorithms, and the performance of the system is improving little by little as the training is deepened. The results of the self-game are also getting better and better, and at the same time, the neural network is becoming more accurate. So, easily defeated the original alpha dog.