<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Omnigrok: Grokking Beyond Algorithmic Data · HackerPapers discussions</title><link>https://hackerpaper.kaipfister.de/p/2210.01117/</link><description>Collected HN comments and submissions linking this paper.</description><language>en</language><lastBuildDate>Mon, 02 Nov 2026 11:16:22 GMT</lastBuildDate><atom:link href="https://hackerpaper.kaipfister.de/p/2210.01117/feed.xml" rel="self" type="application/rss+xml" /><item><title>Omnigrok: Grokking Beyond Algorithmic Data — mentioned by cmorez</title><link>https://news.ycombinator.com/item?id=34476149</link><guid isPermaLink="false">urn:hackerpapers:mention:2210.01117:34476149</guid><pubDate>Sun, 22 Jan 2023 10:55:18 GMT</pubDate><description>&lt;p&gt;&amp;gt; Why do people say they don&amp;#x27;t now how it works &amp;quot;Curve Fitting&amp;quot; is the objective, the function encoded in the weights is the solution, and not actually well understood. See work from Anthropic[1] and Google[2] that explores this. As an analogy consider applying the same argument to the AlphaGo value function. It&amp;#x27;s &amp;quot;just&amp;quot; fitting a bunch of curves to the statistics of millions of self-played games. However, to effectively capture those statistics the network needed to develop a bunch of heuristics. Needless to say these heuristics are not understood (else we&amp;#x27;d already know the principles…&lt;/p&gt;&lt;p&gt;&lt;a href="https://hackerpaper.kaipfister.de/p/2210.01117/"&gt;Paper and discussion history&lt;/a&gt;. A mention is not an endorsement.&lt;/p&gt;</description><category>cs.LG</category></item></channel></rss>
