https://hackerpaper.kaipfister.de/p/2210.16859/Collected HN comments and submissions linking this paper.enMon, 02 Nov 2026 11:16:22 GMThttps://news.ycombinator.com/item?id=46999857urn:hackerpapers:mention:2210.16859:46999857Fri, 13 Feb 2026 07:16:12 GMT<p>Fine, that&#x27;s fair. I believe the statement that you made is countered by my claim, which is: Theorem. For any tolerance epsilon &gt; 0, there exists a transformer neural network of sufficient size that follows, up to the factor epsilon, the policy that most optimally achieves arbitrary goals in arbitrary stochastic environments. Proof (sketch). For any stochastic environment with a given goal, there exists a model that maximizes expected return under this goal (not necessarily unique, but it exists). From Solomonoff&#x27;s convergence theorem (Theorem 3.19 in [1]), Bayes-optimal predictors under the…</p><p><a href="https://hackerpaper.kaipfister.de/p/2210.16859/">Paper and discussion history</a>. A mention is not an endorsement.</p>cs.LG