Top all-time · #773
arg min
Personal blog · argmin.netaimathprogramming
arg min is a personal blog at argmin.net about ai, math and programming. Since 2016, 85 different Hacker News readers have linked it 118 times, most often “Lessons from Optics, the Other Deep Learning”. It is still writing: the last post was on 2026-10-30, and you can follow it by feed.
Sometimes recommended · 9 of 33 sampled comments recommend it?
Why HN links it
Ben Recht's blog on optimization, reinforcement learning, statistics and the critique of ML hype. Commenters cite and recommend specific posts on RL, effect sizes, and mask studies.
- Recommended for · 9 comments
- Commenters recommend posts on reinforcement learning and optimal control, the co-employment post, and the Jacobian/lore-laundering post as worth reading.
Summarised by Claude from a sample of 33 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
arg min: a blog of minimum value. on the history, foundations, and validity of "optimally" automated decision making. Click to read arg min, by Ben Recht, a Substack publication.
Most-linked posts?
- Lessons from Optics, the Other Deep Learning8 people · 9 mentions · first linked 2018-01-25
- The Policy of Truth: Efficiency of Policy Gradients5 people · 5 mentions · first linked 2018-02-20
- Clues for Which I Search and Choose4 people · 4 mentions · first linked 2018-03-20
- Effect size is significantly more important than statistical significance4 people · 4 mentions · first linked 2021-09-14
- An Outsider's Tour of Reinforcement Learning3 people · 4 mentions · first linked 2018-07-22
- The Saga of Highleyman's Data3 people · 3 mentions · first linked 2021-10-22
- Reflections on Random Kitchen Sinks (NIPS “Test of Time” Award Speech)3 people · 3 mentions · first linked 2017-12-12
- Are Developers Out of a Job?3 people · 3 mentions · first linked 2025-07-12
- A look at the math that OpenAI claims computes faster Fourier transforms3 people · 3 mentions · first linked 2026-10-12
- You Cannot Serve Two Masters: The Harms of Dual Affiliation3 people · 3 mentions · first linked 2018-08-10
What Hacker News says?
ali rahimi alludes to the problem of google engineers simply needing to tweak models that were previously tuned by google researchers who do have well-developed intuition [0]. because the intuitions in explicit form are at best heuristic and not necessarily even consistent, signing up to improve a model without them might result in spending indefinite time and compute resources without guarantee of positive results.…
Comment linking /2018/01/25/optics/
on “Machine Learning Crash Course”Submitted Lessons from Optics, the Other Deep Learning as a story · 531 points
HN discussionBen Recht [0] has worked on control theory and reinforcement learning. See his blog post "What we've Learned to control" [1] and the survey paper "A Tour of Reinforcement Learning: The View from Continuous Control" [2] may be of interest. [0]: https://people.eecs.berkeley.edu/~brecht/ [1]: https://www.argmin.net/2020/06/29/tour-revisited/ [2]: https://arxiv.org/abs/1806.09460
Comment linking /2020/06/29/tour-revisited/
on “Algorithms for Decision Making”Submitted Effect size is significantly more important than statistical significance as a story · 376 points
HN discussionYes, I am aware, I did not mean random search as in random actions, but random search with improved heuristics to find a policy. The point being that that the bells and whistles of PPO and other relatively complaticated algorithms (e.g. Q-PROP), namely the specific clipped objective, subsampling, and a (in my experience) very difficult to tune baseline using the same objective, do not significantly improve over grad…
Comment linking /2018/02/20/reinforce/
on “OpenAI Five”Submitted You Cannot Serve Two Masters: The Harms of Dual Affiliation as a story · 255 points
HN discussionIf anyone enjoyed this article, then i'd recommend reading this one as well[1], it's an interesting article with a focus on the relationship between radio and probabilistic reasoning in the early 1900s. https://www.argmin.net/p/the-spirit-of-radio
Comment linking /p/the-spirit-of-radio
on “Radios, how do they work?”Submitted A look at the math that OpenAI claims computes faster Fourier transforms as a story · 249 points
HN discussion> It's a really fabulous study... Ehhhh... not so much. It had serious design flaws in both the protocol and the analysis. This blog post is a fairly approachable explanation of what's wrong with it: https://www.argmin.net/p/are-developers-finally-out-of-a-job
Comment linking /p/are-developers-finally-out-of-a-job
on “AI slows down open source developers. Peter Naur can teach us why”Submitted Prediction Games as a story · 69 points
HN discussionIn case you'd rather read text than watch a video, I believe this is the same? http://www.argmin.net/2017/12/05/kitchen-sinks/
Comment linking /2017/12/05/kitchen-sinks/
on “Greedy, Brittle, Opaque, and Shallow: The Downsides to Deep Learning”Submitted Is the Reproducibility Crisis Reproducible? as a story · 67 points
HN discussionThis article is a little too glib in my opinion, preferring citations and statements to substance and explanations. For a more cutting and insightful critique, watch Ali Rahimi's short talk at NIPS 2017 (where he was presenting for a paper that won the "Test of time" award, for standing out in value a decade after publication). The standing ovation he received at the end indicate that his comments resonated with a s…
Comment linking /2018/01/25/optics/
on “Greedy, Brittle, Opaque, and Shallow: The Downsides to Deep Learning”Submitted The Fall of Big Data as a story · 35 points
HN discussionThe Bangladesh study failed to produce a significant result when simpler nonparametric statistics were used to analyze the data. There were also significant discrepancies in the follow up rate between the control and treatment groups. And even if you ignore those issues, the absolute effect size (20 cases) was tiny. The effect size was expressed in terms of the relative rate of symptomatic seropositives, i.e., norma…
Comment linking /2021/12/01/unblinding/
on “Switching to an N95 mask gives a 75x boost in Covid protection”Submitted Should you go for 2 down 8? as a story · 33 points
HN discussionHere is Ben Recht’s response: https://www.argmin.net/p/toward-a-transformative-hermeneutic...
Comment linking /p/toward-a-transformative-hermeneutics
on “Yes, we did discover the Higgs”Submitted Revisiting the Bangladesh Mask RCT as a story · 9 points
HN discussionThe article here is responding to an original blog post [1] that is not really saying the Higgs was not discovered (despite its trolling title), but raising questions about the meaning of "discovery" in systems that are so complicated as those in modern particle physics. I think the author is using the original motivation of musing on null hypotheses to derive the title "The Higgs Discovery Did Not Take Place", and…
Comment linking /p/the-higgs-discovery-did-not-take
on “Yes, we did discover the Higgs”Submitted Faster Fourier Transforms as a story · 6 points
HN discussionRead https://www.argmin.net/p/lore-laundering-machines Given time, we may find out that the solutions in this paper were also in the literature, as was the case in the anecdotes from the linked article :)
Comment linking /p/lore-laundering-machines
on “Mathematical exploration and discovery at scale”Submitted Overfitting to Theories of Overfitting as a story · 6 points
HN discussionthis is a followup post i wrote to answer exactly your question: http://www.argmin.net/2018/01/25/optics/
Comment linking /2018/01/25/optics/
on “Machine learning has become alchemy (2017) [video]”Submitted Are Developers Out of a Job? as a story · 4 points
HN discussionAdditional context from Ali Rahimi and Ben Recht: http://www.argmin.net/2017/12/11/alchemy-addendum/
Comment linking /2017/12/11/alchemy-addendum/
on “Machine learning has become alchemy (2017) [video]”Submitted Machine learning is not nonparametric statistics as a story · 4 points
HN discussionMake that zero. To their credit, the authors of the Bangladesh mask study released the raw data. "The difference between the two groups was small: only 20 cases out of over 340,000 individuals over a span of 8 weeks." Drawing any conclusions from such small amount of data is deeply unserious. http://www.argmin.net/2021/11/23/mask-rct-revisited
Comment linking /2021/11/23/mask-rct-revisited
on “New Concerning Variant: B.1.1.529”Submitted An Outsider's Tour of Reinforcement Learning as a story · 4 points
HN discussionAlas even today there is no RCT on masks. Why? Countries and states with strict mask policies did not do better than those without these policies (see US States for an example). So do mask work? It is like a religious belief system, you would get the same reaction when questioning if God existed and asked parishioners and atheists. Everything is evidence for those that want to believe. Studies that claim to "prove"…
Comment linking /2021/09/13/effect-size/
on “Viruses that were on hiatus during Covid are back, behaving in unexpected ways”Submitted An Outsider's Tour of Reinforcement Learning as a story · 4 points
HN discussion
