Reinforcement Learning is another large area of AI which has seen a lot of impressive recent developments.

It was at the core of AlphaGo -- which from the AI and Games chapter you already know, and is being applied to various real-world problems, including protein folding, with great success.

Reinforcement Learning also contributes heavily to the success of modern systems such as ChatGPT, Claude, Perplexity, Gemini and others. Part of the success is due to the Large Language Model trained on huge amounts of data from the web and other sources. But perhaps an equal part is due to reinforcement learning helping refine those models so they can follow instructions, do some level of reasoning, use other software tools and so on.

Reinforcement Learning is also used heavily in robotics, for instance, teaching bipedal robots how to balance and walk. Or, teaching your robot vacuum how to navigate your house. So it's definitely worth spending the time digging in and understanding it!

This is a fun chapter, as one of the classic introduction problems for introducing Reinforcement Learning is navigating a maze. So have fun watching our little animated robot try to reach the goal and avoid the traps!