<?xml version="1.0" encoding="utf-8"?>
<rss version="2.0" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom">
    <channel>
        <title>AI kill switch is not the solution – better training is | You&amp;AI</title>
        <link>https://video.echelon.pl/videos/watch/bef4e5c1-7d65-4043-963d-55fa7c6d6e74</link>
        <description>People doing AI development are convinced the technology will kill us all by the end of the decade – these words were on everybody's mind for the past week. Apart from an obvious question of whether this is actually possible, there’s another issue, namely how to prevent that from happening. But as Aleksandra Przegalińska, Vice-Rector at Kozminski University in Warsaw, Poland, tells You&amp;AI’s host Ayşe Abacı, it is not as easy as implementing an “AI kill switch.” More research must be done to understand what’s going on under the hood of AI agents, how they behave in groups, and to better train them – or rather bring them up better. 🔴 Watch our 24/7 livestream - https://youtube.com/live/9RY46AUtIl4 Bringing you all the latest daily news and updates, TVP World is Poland's first English-language channel where you can find world news as seen from the Polish perspective and the latest news from the CEE region. Follow us on Twitter, Facebook, Instagram. https://tvpworld.com/ https://www.facebook.com/tvpworldcom https://twitter.com/TVPWorld_com https://www.instagram.com/tvp_world/ https://www.threads.net/@tvp_world https://bsky.app/profile/tvpworld.bsky.social https://www.tiktok.com/@tvpworld.com https://t.me/tvp_world #AI #Tech #Future</description>
        <lastBuildDate>Mon, 21 Sep 2026 01:40:16 GMT</lastBuildDate>
        <docs>https://validator.w3.org/feed/docs/rss2.html</docs>
        <generator>PeerTube - https://video.echelon.pl</generator>
        <image>
            <title>AI kill switch is not the solution – better training is | You&amp;AI</title>
            <url>https://video.echelon.pl/client/assets/images/icons/icon-1500x1500.png</url>
            <link>https://video.echelon.pl/videos/watch/bef4e5c1-7d65-4043-963d-55fa7c6d6e74</link>
        </image>
        <copyright>All rights reserved, unless otherwise specified in the terms specified at https://video.echelon.pl/about and potential licenses granted by each content's rightholder.</copyright>
        <atom:link href="https://video.echelon.pl/feeds/video-comments.xml?videoId=bef4e5c1-7d65-4043-963d-55fa7c6d6e74" rel="self" type="application/rss+xml"/>
        <item>
            <title><![CDATA[AI kill switch is not the solution – better training is | You&AI - compoundlabs]]></title>
            <link>https://video.echelon.pl/w/pzDA47KDa6btu5nqDod32j;threadId=651</link>
            <guid>https://video.echelon.pl/w/pzDA47KDa6btu5nqDod32j;threadId=651</guid>
            <pubDate>Thu, 17 Sep 2026 22:54:16 GMT</pubDate>
            <content:encoded><![CDATA[<p><span><a href="https://video.echelon.pl/a/kravietz/video-channels" class="u-url mention" target="_blank" rel="noopener noreferrer">@<span>kravietz</span></a></span> A kill switch handles the incident. Training needs a post-deployment test that checks the same request when it is paraphrased, shortened, or split across turns.</p>]]></content:encoded>
            <dc:creator>compoundlabs</dc:creator>
        </item>
    </channel>
</rss>