DeepMind to Teach AI to Play Quake 3 Like a Human

The authors chose "Flag Capture" for AI agents to learn the mechanics of the game in a procedurally-generated world
06 July 2018   520

Researchers from DeepMind, a division of Alphabet in the field of studying AI, reported about a new development designed to teach AI better to play video games. This time, the experts modified Quake III Arena and its "Capture Flag" mode and forced AIs to learn this game.

Mastering the strategy, tactical understanding, and team play involved in multiplayer video games represents a critical challenge for AI research. Now, through new developments in reinforcement learning, our agents have achieved human-level performance in Quake III Arena Capture the Flag, a complex multi-agent environment and one of the canonical 3D first-person multiplayer games. These agents demonstrate the ability to team up with both artificial agents and human players.

DeepMind Team

The authors chose "Flag Capture" for AI agents to learn the mechanics of the game in a procedurally-generated world. Agents played both alone and gathering in teams, including with people. In addition, AI has learned to use tactics such as protecting the base, waiting for the enemy and following the partner:

AI Behaviour
AI Behaviour

The developers used the training method with reinforcement, and the AI ​​did not receive any additional information, except for the picture on the screen. The team of agents trained with each match, receiving a positive response when winning. At the same time, each of them had its own internal reward. AI is based on a pair of recurrent neural networks, fast and slow, each of which studies the transition from scoring points to an internal reward.

According to the results of the study, the authors found that AI agents not only won more often than people, but also were more united. According to the received data, the AI ​​coefficient Elo, responsible for the chance of winning, is higher than the human:

Performance Graph
Performance Graph

The authors claim that in the future they will develop technologies of simultaneous training with the reinforcement of several AI agents, and will also pay more attention to uniting agents and people in teams for greater efficiency. 

AI to Predict Parkinson

Looks like artifical intelligece can be used for really important things
16 November 2018   51

In Oxford, an AI-framework for the diagnosis of nystagmus is created - an early symptom of neurodegenerative pathologies, such as Parkinson's disease. Nystagmus is a form of sleep disturbance, a series of involuntary rapid tremors in the eyeballs of a sleeping person. Rapid diagnosis of nystagmus will allow to treat Parkinson’s disease at an early stage.

The researchers used data from 53 patients from an open laboratory database of the Montreal Sleep Research Archive. Records of electrical activity of the brain, skeletal muscles and eye movements were processed using the algorithm of regression decision trees (random forest).

As the main symptom of nystagmus and the approaching Parkinson's disease, researchers considered muscle atony. In total, electrograms identified 156 different features that can indicate the development of pathology.

Scientists used manual and automatic markup methods for a data set. With manual marking, they managed to achieve diagnostic accuracy of 96%, with automatic results being 4% worse. The researchers plan to improve the results of automatic processing using mathematical functions that mimic the behavior of brain neurons.

A month before the publication of the work of experts at Oxford University, scientists from the Swiss Institute of IRIS reported on the results of work on their own system for diagnosing neuropathology. The fundamental difference is that the Swiss system uses data collected using a smartphone, and the development from Oxford relies on special medical tests.