Angela Collier
2 HN recommenders · recommended in 4 months, 2024–2026
50:06 · 482K views · Jun 18, 2023
Physicist Angela Collier uses Gell-Mann amnesia to critique pop-science figures, especially Michio Kaku, and works through an eigenvalue problem along the way. It is a long, rambling commentary on how experts see simplified explanations of their own field.
HN commenters recommend it for the Gell-Mann amnesia framing and as a skeptical take on Michio Kaku.
5 comments from Hacker News
A couple paragraphs in and I’m already like “Ah, here comes the Michael Crichton quote, the “papers’ full of ‘em” one.” I’ve never read one of these that didn’t come off as unbearably smug. Like sure, discerning the truth is hard, duh. But the people who try their best to do it, even if they sometimes fail, are worth immeasurably more than the dead-souled people who decide “might as well sling bullshit since it’s impossible to be totally correct” (if that’s not what the article is saying out loud, it’s certainly what a lot of the commenters are saying) Edit: Just to end on a positive note, I want to call out the one good piece of content in this genre, which takes Gell-Mann amnesia, flips it, inverts it, and then skewers Michio Kaku for trying to act like he knows about hurricanes: this video
Dr Angela Collier has this fun concept she calls Mann-Gell Amnesia about how basic explanations of topics you are an expert in always seem very unsatisfactory, even when they come from good sources. It's definitely a weird little bias I've noticed in myself! this video
A really useful meta-question is: How much work is it to answer that question? A few examples: Junior devs asking about how to do something - sometimes the answer is clear, there's a standard or a decision has already been made - sometimes someone has to do some work to decide how to proceed. Its easy for people to get frustrated when they think they asked a simple question, but at the point they ask, NO ONE knows the answer. Do masks work? How much evidence of what types would it take to make a satisfying answer to this question? What would those study designs look like? How large would your sample size need to be? I think it would take a great deal of work just to design the studies and get them past ethical review boards, much less get them funded and completed. Anything related to diet. Getting real answers takes way more work than can reasonably be done for anything more that the simplest questions. --- Related: this video
Angela Collier (a theoretical physicist) is also pretty harsh on him [1]. The main critique of him is that he happily keeps talking authoritatively and confidently about scientific topics that he has very little knowledge about (e.g., quantum computing). [1] this video
Ah, yes. The "Igon Value Problem": this video I was thinking the cautiousness was intentionally bred into LLMs, but it could very well be an artifact of the training data, too.