Top all-time · #178
Rich Sutton
Personal blog · incompleteideas.netaiscience
Rich Sutton's personal blog at incompleteideas.net covers ai and science. Since 2016, 240 different Hacker News readers have linked it 297 times, most often “The Bitter Lesson”. No feed was found, so read it on the site itself.
Sometimes recommended · 8 of 40 sampled comments recommend it?
Why HN links it
HN readers link Rich Sutton's site for "The Bitter Lesson" essay, which argues that general methods scaling with compute beat hand-crafted human knowledge. It is also linked for his free Sutton and Barto reinforcement learning textbook.
- Recommended for · 8 comments
- Commenters call the Bitter Lesson essay wonderful and well worth a read, and the Sutton and Barto book beautifully written, freely available and recommended for
Summarised by Claude from a sample of 40 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Most-linked posts?
- The Bitter Lesson165 people · 193 mentions · first linked 2019-03-14
- The Bitter Lesson (2019)23 people · 35 mentions · first linked 2019-03-27
- Reinforcement Learning: An Introduction (2018)15 people · 15 mentions · first linked 2018-02-27
- /3 people · 3 mentions · first linked 2018-06-07
- Reinforcement Learning: An Introduction, Second Edition3 people · 3 mentions · first linked 2018-11-27
- Reinforcement Learning: An Introduction (2018) [pdf]3 people · 3 mentions · first linked 2019-01-16
- Speculations Concerning the First Ultraintelligent Machine (1966) [pdf]3 people · 3 mentions · first linked 2022-03-04
- New Draft of Sutton and Barto: Reinforcement Learning (Nov. 5)2 people · 2 mentions · first linked 2017-10-18
- /book/the-book-2nd.html2 people · 2 mentions · first linked 2019-02-23
- /book/bookdraft2017nov5.pdf2 people · 2 mentions · first linked 2018-05-20
What Hacker News says?
When they say "the start", I think they mean the start of the current LLM era (circa 2017). The main story of this time has been a rejection of the idea that major conceptual breakthroughs and complex architectures are needed to achieve intelligence. Instead, it's better to focus on simple, general-purpose methods that can scale to massive amounts of data and compute (i.e. the Bitter Lesson [1]). [1] http://www.inco…
Comment linking /IncIdeas/BitterLesson.html
on “Genie 3: A new frontier for world models”Submitted Reinforcement Learning: An Introduction (2018) [pdf] as a story · 205 points
HN discussionIn 1997, machines defeated a World Chess Champion for the first time, using brute-force "dumb search." Critics noted that while "dumb search" worked for chess, it might not necessarily be a general strategy applicable to other cognitive tasks.[a] In 2016, machines defeated a World Go Champion for the first time, using a clever form of "dumb search" that leverages compute, DNNs, reinforcement learning (RL), and self-…
Comment linking /IncIdeas/BitterLesson.html
on “AI solves International Math Olympiad problems at silver medal level”Submitted The Bitter Lesson as a story · 152 points
HN discussionIf you haven’t already read it, then you may find “The Bitter Lesson” essay interesting to read. http://www.incompleteideas.net/IncIdeas/BitterLesson.html
Comment linking /IncIdeas/BitterLesson.html
on “Sharing AI progress in mathematics”Submitted Reinforcement Learning: An Introduction (2018) as a story · 120 points
HN discussionBy all means if it works to solve your problem, go ahead and do it. The reason some people have mixed feelings about this because of a historical observation - http://www.incompleteideas.net/IncIdeas/BitterLesson.html - that we humans often feel good about adding lots of hand-coded smarts to our ML systems reflecting our deep and brilliant personal insights. But it turns out just chucking loads of data and compute a…
Comment linking /IncIdeas/BitterLesson.html
on “Claude 3 model family”Submitted The Bitter Lesson (2019) as a story · 98 points
HN discussionPSA: The new edition of Sutton/Barto has a nice discussion of (the original) AlphaGo in the back. http://incompleteideas.net/sutton/book/the-book-2nd.html
Comment linking /sutton/book/the-book-2nd.html
on “AlphaGo Zero: Learning from scratch”Submitted The Bitter Lesson (2019) as a story · 84 points
HN discussionWow, that was a really deep and enjoyable Wikipedia rabbit hole journey. I hadn't heard of Temporal Difference before (though I was familiar with Q-learning). It was interesting to note that TD-Gammon improved with expert designed features. I wonder if this was simply related to the technology of the field as it stood over 20 years ago or some underlying categorization or complexity associated with the games themsel…
Comment linking /sutton/
on “AlphaGo Zero: Learning from scratch”Submitted The Bitter Lesson (2019) as a story · 45 points
HN discussionIn machine learning, hands down these are some of the best related textbooks: - [0] Pattern Recognition and Machine Learning (Information Science and Statistics) and also: - [1] The Elements of Statistical Learning - [2] Reinforcement Learning: An Introduction by Barto and Sutton - [3] The Deep Learning by Aaron Courville, Ian Goodfellow, and Yoshua Bengio - [4] Neural Network Methods for Natural Language Processing…
Comment linking /book/bookdraft2017nov5.pdf
on “Ask HN: What are the best textbooks in your field of expertise?”Submitted Reinforcement Learning: An Introduction, Second Edition as a story · 39 points
HN discussion- nl · 2026-08-28
> a large language model trained on chess commentary as well as being trained on millions of chess games would outperform stockfish which is only trained on millions of chess games Not really, if anything it's closer to the opposite. The Bitter Lesson essay literally has this as an example: > These researchers wanted methods based on human input to win and were disappointed when they did not.[1] and > Enormous initi…
Comment linking /IncIdeas/BitterLesson.html
on “Small Models Have Arrived” Submitted Verification, the Key to AI (2001) as a story · 36 points
HN discussionThis idea has failed to pan out time and time again - people have an instinct that hand-crafted finely-tuned specialized AI systems must be optimal, but throwing more scale and compute to something more generally smart always wins out. It's especially palpable just looking at the last few years of LLM's: a frontier model with all the world knowledge you can stuff in it and every tool at its disposal has always perfo…
Comment linking /IncIdeas/BitterLesson.html
on “Small Models Have Arrived”Submitted Bitter Lesson by Richard Sutton as a story · 12 points
HN discussionI can see both sides of the argument. On one side, using ML feels like huge overkill when a simple trick exists. Plus AI can freak out in some circumstances. On the other side, it may find other, less obvious cues giving something more robust. Rich Sutton's "bitter lesson" says the weight will move in time in favor of ML. http://www.incompleteideas.net/IncIdeas/BitterLesson.html
Comment linking /IncIdeas/BitterLesson.html
on “The First Rule of Machine Learning: Start Without Machine Learning”Submitted The Bitter Lesson as a story · 5 points
HN discussionThis book seems to focus on large language models, for which RLHF is sometimes a useful addition. To learn more about RL, most people would advise the Sutton and Barto book, available at: http://incompleteideas.net/book/the-book-2nd.html
Comment linking /book/the-book-2nd.html
on “Implementing a ChatGPT-like LLM from scratch, step by step”Submitted Speculations Concerning the First Ultraintelligent Machine (1966) [pdf] as a story · 4 points
HN discussionMaybe, if you want to relearn the bitter lesson. http://www.incompleteideas.net/IncIdeas/BitterLesson.html
Comment linking /IncIdeas/BitterLesson.html
on “OpenAI, Google and Anthropic are struggling to build more advanced AI”Submitted The Bitter Lesson as a story · 4 points
HN discussion> Since 3.5-sonnet, we have been monitoring AI model announcements, and trying pretty much every major new release that claims some sort of improvement. Unexpectedly by me, aside from a minor bump with 3.6 and an even smaller bump with 3.7, literally none of the new models we've tried have made a significant difference on either our internal benchmarks or in our developers' ability to find new bugs. This includes th…
Comment linking /IncIdeas/BitterLesson.html
on “Recent AI model progress feels mostly like bullshit”Submitted Reinforcement Learning: An Introduction [eBook, Code, Etc] as a story · 4 points
HN discussionThis article conflates two separate, very different issues into one: * Issue #1: There is a tremendous amount of hype, noise, and snake oil surrounding the moniker "AI." Pretty much everyone agrees with this statement. (And anyone who doesn't agree with it... is probably selling snake oil.) * Issue #2: Is intelligence just a form of "curve fitting," i.e., is it just finding solutions to very complicated, high-dimens…
Comment linking /IncIdeas/BitterLesson.html
on “AI is mostly about curve fitting (2018)”Submitted The Bitter Lesson as a story · 3 points
HN discussionThe above is a common anthropocentric take that has been repeatedly disproven by the last decade of deep learning research: http://www.incompleteideas.net/IncIdeas/BitterLesson.html Understanding audio with inputs in the frequency domain isn’t required for understanding frequencies in audio . A large enough system with sufficient training data would definitely be able to come up with a Fourier transform (or somethin…
Comment linking /IncIdeas/BitterLesson.html
on “Vector Databases: A Technical Primer [pdf]”Submitted The Bitter Lesson as a story · 3 points
HN discussion
