11 comments

  • vintermann3 hours ago
    From my experience, modern machine learning models don&#x27;t scale well <i>down</i> at all, and it&#x27;s almost certainly better to just use a simple Markov chain variant of some sort - like Niall on the Amiga, or whatever Terry Pratchett used to come up with Foul Ol&#x27; Ron&#x27;s catchphrase, &quot;Millennium hand and shrimp&quot;.
  • derefr8 hours ago
    &gt; The model weights and inference code need to be contained within 25KB of user-space memory<p>Wouldn’t it be era-appropriate to allow relying on banked memory? You’d still need to hold the inference code, but you could effectively stream(ing page) the weights as you compute on them.
    • striking6 hours ago
      From a look at <a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;BBC_Micro#Specifications" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;BBC_Micro#Specifications</a> I think the 6502 versions of the beeb didn&#x27;t have banked RAM so to keep it loadable from tape the limits might be as stated.<p>But with substantial additional effort, maybe some banked ROMs could be added..?
      • froh4 hours ago
        the banking was done via an external ram controller whose bank control register was mapped to some (unbanked) controller address.
  • bmc750510 hours ago
    Cool to think this demo would have been possible over fifty years ago. I wonder what someone from 1975 would have said if you had shown this to them back then.
  • tyromaniac10 hours ago
    This is super cool! As someone who&#x27;s worked a little with NES programming and tried out cc65, I&#x27;m surprised he didn&#x27;t just hand write some assembly, he likely couldve saved a lot of space if I had to guess.
  • actionfromafar10 hours ago
    The 6502 is notoriously unfit for a C compiler, so probably there is room for more performance in the future. :)
  • torment-nexus7 hours ago
    The biggest win for AI dev efficiency is cutting down what gets loaded into context. Semantically matching tasks to the top tools helps a lot.
  • toplinesoftsys10 hours ago
    This is amazing project! I hope it will result in real miniaturization of AI - for example, edge LLM inside of glasses. That will be awesome.
  • aghilmort5 hours ago
    really great work
  • leonmeng6 hours ago
    [flagged]