‹ BackHN Continuity

Thread

Can gzip be a language model?

414 points · 165 comments · networked

  1. adamgordonbell · · focus · HN ↗
    Compressing something and next token prediction are very related. Once you understand the connection, things like ts_zip and hutter prize make a lot more sense.

    <a href="https:&#x2F;&#x2F;bellard.org&#x2F;ts_zip&#x2F;" rel="nofollow">https:&#x2F;&#x2F;bellard.org&#x2F;ts_zip&#x2F; <a href="https:&#x2F;&#x2F;corecursive.com&#x2F;the-hutter-prize&#x2F;" rel="nofollow">https:&#x2F;&#x2F;corecursive.com&#x2F;the-hutter-prize&#x2F;

Open on Hacker News to reply ↗

Unofficial Hacker News client; not affiliated with Y Combinator.