Language Models can Solve Computer Tasks · HackerPapers discussionshttps://hackerpaper.kaipfister.de/p/2303.17491/Collected HN comments and submissions linking this paper.enMon, 02 Nov 2026 11:16:22 GMTLanguage Models can Solve Computer Tasks — mentioned by bob1029https://news.ycombinator.com/item?id=35392359urn:hackerpapers:mention:2303.17491:35392359Fri, 31 Mar 2023 19:39:33 GMT<p>The problem with this kind of writing is that the AI space is moving so quickly that some of these criticisms were literally addressed yesterday. https://arxiv.org/abs/2303.17491 &gt; In this work, we show that a pre-trained large language model (LLM) agent can execute computer tasks guided by natural language using a simple prompting scheme where the agent recursively criticizes and improves its output (RCI). I also agree with the GPT3.5 vs 4 comparison made by other comments here. Arithmetic and other emergent phenomena only begin to take off beyond certain parameter counts. Stuff as basic as…</p><p><a href="https://hackerpaper.kaipfister.de/p/2303.17491/">Paper and discussion history</a>. A mention is not an endorsement.</p>cs.CL