AI Learned to Play Hide-and-Seek Using Machine Learning

OpenAI announced its new project in which an AI can learn to play hide and seek. Utilizing the Machine learning language, you can play hide and seek with AI. Initially, a simple version of the game will be released in which ‘Seekers’ will score when a ‘Hider’ is visible to the field. At the start of the game, the ‘Hider’ will be given some time to set up a place for hiding.

To add the pinch of interest and fun in the game, ‘Seekers’ and ‘Hiders’ both can move any object over the field say walls and blocks, to take a lead over others. The game might sound simple but the results are impressive. Out of the heap of 481 million games of hide-and-seek, this AI-based game can frame strategies and counterstrategies. Therefore, simple physics, cool graphics, and logical aspects make it quite appealing.

OpenAI has shown that working with even simple machine learning and AI technique, can result in developing amazing apps like virtual girlfriend simulator. In such games, AI plays showing unexpected behavior like humans, where it gets rewards for a good move. In short, it’s all about the application of Reinforcement learning. Taking here the example of a game – Hide-and-seek to analyze the correct application and results obtained by AI applications.
Let’s have a look at some of the highlights of the game. –

Running, Chasing, and Hide

There are many Hide and Seek games out there but this one seems very natural playing with AI where AI agents will run and chase the one who is hiding on the map.

Building Place To Hide

AI agents can lock the blocks in place and the one who can lock it only can unlock it. AI agents have to build a place to hide through blocks available.

Ramps To Break Into The Shelter

The game has ramps. Using these ramps, Seekers can easily break into the shelter and find the hider. However, hiders have also learned to counterattack. Hiders can freeze the shelter openings to freeze the seekers inside. They thought that it could be the end of the game, but OpenAI stated, “That’s wrong”.

Use Box to Hit The Shelter

Using the box, they can hit the shelter and jump into the shelter through it. In this way, they once again find the ‘Hider’.

This is how the AI Agent plays Hide-and-seek. Agents will be rewarded when they finally win the game. This is one of the good and impressive games using the technique of reinforcement learning. The game shows strategic decisions with the simple use of machine learning and AI applications. With the help of reinforcement learning, we can do a lot of things that make our life easier.

As we’ve witnessed a lot of unexpected behavior in the game of Hide-and-seek. On the other hand, when we have such surprising elements of reinforcement learning, there could be some hazardous and undesired behavior as well.
So, As AI is growing with each passing day, we also have to move along with the dynamic AI to ensure that they deliver only what we want and nothing undesirable.