The first human-AI match in Vietnam

The final round of "AI Arena" – a gold mining game simulation competition in Esports form will take place on September 26^th.

AI- Reinforcement Learning arena, the first contest for Vietnamese technology engineers who are passionate about artificial intelligence (AI) and Reinforcement Learning has been organized by FPT Software on the CodeLearn programming platform since August 8th, 2020. The contest attracted nearly 1,000 Vietnamese contestants divided into 445 teams from Vietnam and many other countries such as Japan, Korea, Germany, America, etc..

In the final round called "Treasure Island", Top 8 teams passing the qualifying rounds will join 15 combats simulating the form of Esport to find the winner. This is the first time there is a game simulation competition using AI algorithms combined with Esport in Vietnam. Watch the final leg on FPT Software fanpage and participate in the minigame with many attractive prizes (up to 1.6 million VND/prize).

Among the teams that entered these final matches, 4 teams created virtual agents based on the Reinforcement Learning algorithm, the remaining 4 teams used a different algorithm from the AI ​​branch to create "gold miners" of theirs. This result will be judged by the experts of the competition as a battle between human intelligence and artificial intelligence.

The special feature of this algorithm is that the participants must program it  so that the agent can learn and make decisions strategically. The virtual machines then have to upgrade their "brains" to self-learn moves, the rules of the game, and develop strategies to beat the opponent.

The gold digging game on CodeLearn platform in "AI Arena"

"Reinforcement learning is a new and exciting field not only in Vietnam but also around the world. Over the past few months, the team has gradually trained its 'gold miner' to become more and more professional thanks to the battles with others and the suggestions from the experts," said the representative of DeepShuttling team from Singapore.

After nearly two months, "The AI ​​Arena" is becoming an intellectual event between humans and artificial intelligence. Engineers had to program their teams’ agents to self-learn and win against the opponents’ AI. This is also the principle of the Reinforcement Learning algorithm: Self-study and experience accumulation. Correct moves are saved by the machine, the ones which are false or do not work are rejected. The effectiveness of the algorithm has been proven by the fact that AI can beat real people in the most difficult games today from eSports, such as Dota2, Starcraft II, to chess (go, chess).

After the group stage, participants must continue to update their programming code with the consultation of leading AI experts, such as Mr. Nguyen Xuan Phong - Mila Institute of Artificial Intelligence, Mr. Vu Huu Tiep - Machine Learning engineer at Google.

"I and the organizers are quite surprised by the quality of the contestants in this competition. Strengthening is a difficult algorithm, while the problem posed by the Organizing Committee is not easy, the teams have competed fiercely to prove their intelligence and ability, creating an attractive group stage. Many teams which achieved impressive breakthroughs by grasping and applying the reinforcement learning algorithm makes 'AI Arena' much more interesting,” said Mr. Nguyen Xuan Phong, an international AI expert, collaborating at the Mila Institute for Artificial Intelligence.

The winner of the AI- Reinforcement Learning competition will receive a prize of VND 100 million and a visit to the Workshop on AI worth VND 20 million organized by FPT Software. In addition, the Organizing Committee will also award "Talent" worth VND 20 million and "Promising" worth VND 10 million to a team with 100% of members being high school students or students from university.