Another striking breakthrough in the field of artificial intelligence (AI) - A new version of the "AlphaGo" computer program can start from a blank state without the need for any human chess game, according to a recent article published in the "Nature" magazine. Under the input conditions, self-study Go, no teacher self-taught, self-taught.
This new program is called AlphaGoZero and it defeated the old Alpha Dog who once wrote history with an unbeaten record of 100:0.
Strength, brilliant achievements of "Alpha Dog"
Artificial intelligence "Alpha Dog" comes from Google's DeepMind company. As early as January 2016, Nature magazine reported on the cover position that the program achieved a 99.8% win rate against all other Go games and defeated the European Go champion with a score of 5..0.
Then in March of the same year, "Alpha Dog" challenged world chess champion Li Shishi. After five rounds of fierce battles, humans 1...4 did not suffer from enemy intelligence, causing a huge sensation. The people's shock stems from the fact that on the Go 19x19 board, there are even hundreds of ways to go. The traditional artificial intelligence algorithm is difficult to estimate the situation and the next.
But "Alpha Dogs" is still accumulating strength. In January of this year, a mysterious account Master announced that he is the new version of the "Alpha Dog". The account on the well-known Go platform has challenged Ke Jie, Pu Tingyi and Jingshan Yutai, and other top masters, hoping to win, and then winning the "Chess" Nie Weiping, until the 60 disk burst identity, finale to close.
In 3 days, the "Alpha Adamoto" self-help
In this latest paper, David Silver, Dymes Hassabis, and his colleagues, who are the main leaders of the London Alpha Dog Project of the London-based Deep Thinking Company, reported on the new Alpha Dog software. - "Alpha Ada."
The most striking thing about the "Alpha Adamoto" lies in its ability to teach itself. Its learning begins with “zero†and is based solely on its own game. It does not require the guidance of any historical record, nor does it require any human experience.
For this new software, human input is limited to only a board and a pawn, and no human data is involved. "Alpha" uses only one neural network. This network is trained to predict the program's own moves and the winners of the game, so that "Alpha Avatar" progresses in each self-trick.
Through three days of training - including nearly 5 million games of self-play - "Alpha Ace" has been able to surpass humans and defeat the previous "Alpha Dog" version. David Silver said that "Alpha Ada" is far more powerful than "Alpha-Dog". It is no longer limited by human knowledge, but can discover new knowledge on its own.
Get rid of the help and dependence of human experience, artificial intelligence has become even stronger! After the "Alpha Affluence" emerged, people speculated: Do we humans also delay the "Alpha Dog"?
Shocked, independent discovery of AI
In fact, one of the biggest challenges of artificial intelligence is the development of an algorithm that can start with “zero†and learn complex concepts at a superhuman level.
In the world-famous game last March, the Alpha Dog software defeated the human Go world champion. At that time, the Go program used value networks to calculate the situation and use strategic networks to choose the next.
However, in order to win this man-machine battle, the team of scientists used supervised learning (based on the steps of chess for millions of human professional players) and reinforcement learning based on self-taught while training “Alpha-Dogâ€. The old Alpha Dog training process lasted several months and used multiple machines and 48 TPUs (professional chips for neural network training).
This time, according to the research team, the new program "Alpha" uses only one machine and four TPUs.
Before Alpha Dog, the most successful Go game can only reach the level of human amateur players, and can't compete with professional players in the absence of children. Originally, people believed that artificial games should be artificial for at least 10 years. Intelligence can achieve this achievement. But the "Alpha Dog" turned out to prophesy. In less than two years since its inception, Alpha Awakening has once again refreshed people's perceptions. With the progress of procedural training, artificial intelligence has independently discovered the rule of Go that humans have used for thousands of years. It has also been established. With the new strategy, it can be said that artificial intelligence has brought new insights into this ancient game.
MTU Gas Generator Powered by MTU gas engine, alternator, radiator, controller, base frame;
. World Famous diesel engine brand: MTU Gas engine
. World famous AC alternator brand: Stamford, Leroy Somer, Mecc Alte, Marathon, Faraday, SWT
. World famous genset controller brand: Deepsea, ComAp, Deif, SmartGen, Motortech
. Gas Control System: Ignition system, Gas Throttle System, Ga Mixer System, Gas Train Valve System
. Start Battery system
. Optional for Remote Cooling system with CHP & CCHP Control
. CHP- Combine with Heat and Power Electrical system
.CCHP- Combine with Cold, Heat and Power electrical system
Gas Engine,Mtu Gas Genset,Mtu Gas Generator,Mtu Biogas Generato
Guangdong Superwatt Power Equipment Co., Ltd , https://www.swtgenset.com