4.2 Snake

At this station, you can challenge three artificial intelligences in a game of Snake.

The goal is simple: eat as many apples as possible without crashing into yourself.

Even if you’ve never played before, a couple of rounds will be enough to understand how to move and score points.

The three artificial intelligences you’re competing against look identical — but they differ in how much training they received: one trained for 100 games, another for 10,000, and the last for 10 million.

But what does “trained” mean?

At first, the AI plays randomly and makes errors that to us seem absurd. Each game generates a score — and that number becomes feedback. Based on results, the system gradually changes its strategy: keeping the decisions that increase the score on average and discarding those that don’t.

After repeating this process many times, the strategy improves — and the AI becomes more and more skilled… until it is nearly unbeatable.