Not a perfect video by any means but I’m going to go ahead and say I no longer recommend Feynman’s books except the lectures on computation. I haven’t reviewed the books and this isn’t an anti-recommendation, just a removal of any positive recommendation. I do think she has some good criticisms about a lot of the stories in Surely You’re Joking not actually being true or being misogynist.
One interesting thing about the video is she complains a lot about Feynman having ghostwriters who aren’t disclosed well enough, not writing his own books. And I think that’s a valid complaint. But I knew Feynman didn’t personally write those books. Tons of other ghostwriting is less disclosed than this. She also complains about the lack of editing of some Feynman books based on audio recordings, but I’d rather have that than significant editing like was done for Ayn Rand Answers.
CF involves connecting ideas with goals and breaking goals down into many simple sub-goals. This makes ideas much easier to evaluate: check if it works for each sub-goal. This lets thinking be more of a flowchart method. There’s still creativity involved, but you get more guidance than unstructured thinking. Induction and MCDM are highly flawed while CF offers something that doesn’t have known dealbreaker flaws. Caring a lot about dealbreaker flaws is also something CF emphasizes more than alternatives. CF says to categorize issues as dealbreakers or not (more often called “decisive” or not) and focus on decisive issues. Instead of one vague high level goal, CF uses many small, simple, decisive, binary sub-goals which allow for much easier, faster objective evaluation. The point is to use ideas (which can be answers, theories, decisions, actions, plans, conclusions, etc) which have no decisive errors. CF rejects indecisive, quantitative reasoning about how good things are and replaces it with looking for what issues will cause failure at what (sub-)goals. CF focuses on failure not success because one error can cause failure while dozens of good traits cannot cause success (the addition of some errors can still lead to failure despite all the good traits). Also error correction correction is how evolution works and contradiction is a powerful logical tool with no equivalent positive version.
omg i was asking multiple AIs today to find typos in a draft and all of them told me typos inside quotations. i said i can’t edit that since it’s not my words and then they agreed we shouldn’t edit it. but by default they tell people to edit quoted text. ughhhh
This looks interesting to me. Maybe important. Not sure.
He says you can replace all buttons on a scientific calculator with only 2 buttons and do all the same things. The buttons are the number 1 and the function he discovered, eml:
seems curious, but also nand is primitive[1], but eml’s definition has:
exp and log (obviously)
addition/subtraction
multiplication (if you don’t count subtraction; use *-1 and + instead)
I think you could make a brute force version of this that uses patterns like (1 - p_1)(\text{expr}) + \cdots to make irrelevant factors 0 (also p_i \in \{0,1\}).
we call it not and but if and is primitive i don’t see why nand shouldn’t be; and is just not nand. ↩︎
Yeah I thought of brute forcing it with e.g. a function that adds every primitive and has separate inputs for all of them and also a coefficient for every term with its own input. Then you can zero out all primitives but the one you’re currently using.
But doing it with only 2 inputs and 3 primitive operators (according to the current standard understanding of what is primitive), and via a single fairly simple combination of them instead of using them independently, seems pretty clever, not brute force.
