Everyone’s talking about AI. But few people know how AIs are trained. That’s what makes this project special. It’s an illuminating explanation of how AIs can be trained, using Pokémon as an example.

The creator, Peter Whidden, is a guy from YouTube who’s great with code and technology. He explained how he trains it by simulating its actions, movements, and success in completing tasks. He even gives virtual rewards based on the AI’s success in catching Pokémon, leveling up, and not dying. By the end of the video, you will understand how computers respond to a task, what challenges them, and how to improve their responses.

For his project, he uses reinforcement learning (a type of machine learning) to train an AI to play Pokémon games. What I love the most about this is that there’s a degree of over-complication that makes it a bit of a satirical usage of the technology. While people are using AI to solve bigger and more complex problems, it can be a tool to experiment and create a conversation.

Overall, the video explains how the technology works – in this case, training an AI to play Pokémon – in a simplified manner that makes the content very interesting, and serious, yet highly credible – a great combination that has led to 6 million views on YouTube alone. He also pays close attention to how he visualizes the data, by overlapping the training data, creating a simplified iconography, and creating a story arc about the AI and its training journey that makes you want to root for its success.