The speaker explains the different types of reinforcement learning, policy-based, model-based, on/off-policy, and value-based, and why it's difficult to apply reinforcement learning in most settings today.