Jan :rust: :ferris:<p><span class="h-card" translate="no"><a href="https://kolektiva.social/@d10c4n3" class="u-url mention" rel="nofollow noopener" target="_blank">@<span>d10c4n3</span></a></span> </p><p>Yes, this is another issue with the current <a href="https://floss.social/tags/LLM" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>LLM</span></a> architecture: they can't learn or improve. It's one big monolith - not composed of smaller parts (that could individually be improved or composed with each other).</p><p>There _are_ AI architectures that are _actually reliable_...</p><p><a href="https://floss.social/@janriemer/114454349034565839" rel="nofollow noopener" translate="no" target="_blank"><span class="invisible">https://</span><span class="ellipsis">floss.social/@janriemer/114454</span><span class="invisible">349034565839</span></a></p><p>...but our industry has decided to hyperfocus on bullshit machines.</p><p><span class="h-card" translate="no"><a href="https://mastodon.online/@larsmb" class="u-url mention" rel="nofollow noopener" target="_blank">@<span>larsmb</span></a></span> </p><p><a href="https://floss.social/tags/ArtificialIntelligence" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>ArtificialIntelligence</span></a> <a href="https://floss.social/tags/FormalVerification" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>FormalVerification</span></a></p>