After artificial intelligence defeated the top human masters in chess, Go and other projects, Allen Musk's OpenAI artificial intelligence beat the top player Dendi in the popular online game "Dota2". On August 12th, Alan Musk issued a message on his twitter account, praising the victory of OpenAI, and said that this is the first time artificial intelligence has defeated world-class players in confrontational e-sports. OpenAI is a non-profit AI research company founded by Tesla CEO Alan Musk, which conducts research on secure artificial intelligence. Previously, Musk's company had successfully trained AI, performed tasks on VR devices, and immediately developed robots to create their own language systems.
OpenAI explained in its blog: "This self-play self-learning of Dota2's game enables artificial intelligence to capture operational performance from a learning system that is less than human, and then leap to full ownership. Superman with computing power. By contrast, Deep-learning deep learning systems can only achieve the same level of intelligence as their training database, but in self-learning systems, when data is upgraded, existing data will automatically improve.
OpenAI competition process
The Dota2 game started in March of this year, and the original AI robot only did some simple games. As a free multiplayer esports game, Dota2 is most proud of not limiting the player's limits. On August 7th, the robot achieved an impressive winning streak and defeated the Dota2 player named Blitz in the first game. On the same day, the robot defeated two other high-rank players. The next day, it defeated another highly respected player, Arteezy. And all four players defeated by the robot agreed that player Sumail can beat it. However, Sumail is not as good as his comrades-in-arms. In the end, the robot challenged Dota2's former world champion Dendi and beat Dendi 2-0. "Please let me go!" Dendi told the robot opponents during the game.
Self-play self-learning process
OpenAI researchers said that in this competition, robots use self-learning from scratch, not imitating learning or tree search, establishing clear goals in a chaotic and complex environment involving humans, and AI system construction has taken a historic step. In the competition, the robot predicts the player who is about to move, and makes a reaction in the new environment and affects other players. Between the two games, OpenAI staff combined some “training†with self-learning to help the robot continuously improve its combat effectiveness.
The goal of OpenAI is to build secure, universal artificial intelligence. For AI, Dota is an ideal test site with intense competition and complex games. If you want to achieve success in this game, you must advance the existing technology and achieve technological breakthroughs. We have now developed an AI that can beat top pros in Dota's 1v1 game.
Dota's rules are very complicated. If you use the rule method, the built AI must be a low-energy player. Therefore, OpenAI completely adopts a self-play training method. At the beginning of training, AI has no knowledge of the world in which he lives, but only competes with his own replicas. This means that it is always equal to its opponents. In this process, self-ability continues to increase until it reaches The level of the world's top players.
Self-learning development
The concept of “self-defeating†is the key to OpenAI's R&D. This is an effective way for AI systems to learn to solve extremely complex tasks: against a weak or too strong opponent, it can't learn from it, but it is a valuable opponent. OpenAI researchers explained: "You can see that AI has evolved from a completely random state to a top level today."
In addition to OpenAI, other artificial intelligence researchers are already looking for ways to teach robots to do everything: from reading to dreaming. They created AI systems that they can learn on their own, and even AI systems that can teach each other. Education is fundamentally subverted to meet the needs of this new technology, and with the wisdom of robots, our lives will become easier, safer, and more interesting.
In addition, Google's Deepmind team recently conducted research to allow AI to "sport" in complex environments. The researchers designed a stickman-like model that required them to cross the obstacle to reach the end without any programming, so the stickman was basically trying to make a mistake. During the experiment, the AI ​​will gradually adjust its joints and accomplish the goal in a better way. The stickman of the sports spirit burst table has 27 free joints and 21 drives. And in order to guide this AI, Deepmind designed a set of virtual sensors to teach them how to pass obstacles in unfamiliar complex environments, and the stickman's successful parkour after repeated falls shows AI's powerful self-learning ability. Achieving a combination of machine learning and systemic nerves is a small step for stickies, but it is a big step for AI in the field of human function research.
We offer wide range of coffee tables which include wood coffee tables ,steel coffee tables, wood coffe tables. We have many coffee tables in different sizes. When you buy Sofa for your living room, coffee tables are necessary.Our customers have been using the coffee table which were made in our factory more than 10 years without any problems. They like it very much and continue to buy the furniture from us after a few years.
Coffee Table,New Design Coffee Table,Contemporary Coffee Table,European Style Coffee Table
DELO SOFA , https://www.modernluxuryfurnitures.com