[HN Gopher] Gwern's "Stem Humor" Directory
       ___________________________________________________________________
        
       Gwern's "Stem Humor" Directory
        
       Author : surprisetalk
       Score  : 23 points
       Date   : 2025-11-22 17:49 UTC (5 hours ago)
        
 (HTM) web link (gwern.net)
 (TXT) w3m dump (gwern.net)
        
       | throwaway314155 wrote:
       | Not directly related, but I'm curious if gwern ever caved on
       | their views about GAN's being "abandoned" for diffusion?
       | 
       | https://gwern.net/gan
        
         | gwern wrote:
         | No. I think the need for adversarial losses in order to distill
         | diffusion models into one-step forward passes has provided
         | additional evidence that GANs were much more viable than
         | diffusimaxis loudly insisted.
         | 
         | (Although I'm not really current on where image generation is
         | these days or who is using GAN-like approaches under the hood
         | or what are the current theoretical understandings of GAN vs AR
         | vs diffusion, so if you have some specific reason I should have
         | "caved", feel free to mention it - I may well just be unaware
         | of it.)
        
           | throwaway314155 wrote:
           | "SotA diffusion uses adversarial methods anyways" seems like
           | a bit of a departure from the case you make in the blog post.
           | 
           | edit: For what it's worth - I agree. At least some auto-
           | encoders (which will produce latents for diffusion models)
           | use some form of adversarial method.
           | 
           | Still, I'm curious if you think GAN models in their more
           | familiar form are going to eventually take on LCM/diffusion
           | models?
        
       | woopwoop wrote:
       | https://people.mpim-bonn.mpg.de/zagier/files/math-mag/63-2/f...
       | 
       | "How often should you beat your kids?", by Don Zagier.
       | 
       | > This note is a follow-up to the note "How to beat your kids at
       | their own game" by K. Levasseur, in which the author proposes the
       | following game to be played against one's own children: ...
       | Levasseur analyzes the game and shows that on average you will
       | have a score of n + (sqrt(pi * n) - 1) / 2 + O(n^{-1/2}), while
       | the kid, of course, will have an average score of exactly n.
       | 
       | > We maintain, however, that only the most degenerate parent
       | would play against a 2-year-old for money, and that our concern
       | should therefore be not by how much you expect to win, but with
       | probability you will win at all.
        
       | GreyZephyr wrote:
       | I really want people to crowdsource the DMT prime factorisation
       | project. I know at least one person tried but lost interest
       | before they met an elf. It just seems like such a fun experiment
       | to run. Is it possible to recall numbers at all while taking DMT?
       | Can you memorise new ones? If not why not etc, and maybe the
       | machine God factorises numbers for you!
        
       ___________________________________________________________________
       (page generated 2025-11-22 23:00 UTC)