Andrew Barto and Richard Sutton1 receive the 2024 Turing award for their groundbreaking work on reinforcement learning. Even with today’s focus on LLMs and ‘generative AI’, reinforcement learning plays an important role, e.g. as the ‘RL’ part of RLHF. I still remember how much I enjoyed their classic textbook on the subject (Reinforcement Learning: An Introduction) as a PhD student; it’s freely available2 and one of my all-time favorite computer science books. If you would rather watch something on the topic, the AlphaGo documentary is a nice showcase on how these ideas (combined with neural networks and lots of compute power) can be successfully applied to problems like Go, a wonderful board game that was previously thought ‘unsolvable’.
On a related note also concerning famous computer science textbooks, here is a nice breakdown from 2021 (via HN) of the time it takes to fully work through Structure and Interpretation of Computer Programs (by Abelson, Sussman, and Sussman), colloquially known as SICP3: roughly 729 hours, or almost 20 weeks of full-time work! 🤯
Also, of “The Bitter Lesson” fame. ↩
This is the ‘new’ second edition that’s already seven years old (the first edition is from 1998). Wow, time flies. ↩