<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:itunes="http://www.itunes.com/dtds/podcast-1.0.dtd" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title><![CDATA[Aligned to Flourish, narrated]]></title><description><![CDATA[Narrated readings of the essays on Aligned to Flourish. Jacques Thibodeau writes about automating alignment research, what models actually know, and the strategy of getting to safe AGI.]]></description><link>https://jacquesthibodeau.com/</link><language>en</language><atom:link href="https://jacquesthibodeau.com/podcast/rss.xml" rel="self" type="application/rss+xml"/><lastBuildDate>Tue, 19 May 2026 00:00:00 GMT</lastBuildDate><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><itunes:type>episodic</itunes:type><itunes:category text="Technology"/><itunes:owner><itunes:name><![CDATA[Jacques Thibodeau]]></itunes:name></itunes:owner><copyright><![CDATA[Copyright 2026 Jacques Thibodeau]]></copyright><item><title><![CDATA[Fooled by Vast Knowledge]]></title><link>https://jacquesthibodeau.com/fooled-by-vast-knowledge/</link><guid isPermaLink="false">https://jacquesthibodeau.com/fooled-by-vast-knowledge/#audio</guid><description><![CDATA[Models look like they generalise because the training data is vast enough to hide the difference between interpolation and extrapolation. I think that confusion is leading safety research astray.]]></description><itunes:summary><![CDATA[Models look like they generalise because the training data is vast enough to hide the difference between interpolation and extrapolation. I think that confusion is leading safety research astray.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/6a0cc9e75ff9f64634b07408/audio.mp3" length="8780198" type="audio/mpeg"/><itunes:duration>9:09</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Tue, 19 May 2026 00:00:00 GMT</pubDate></item><item><title><![CDATA[When Execution Gets Cheap, Does Taste Become the Moat?]]></title><link>https://jacquesthibodeau.com/when-execution-gets-cheap-does-taste-become-the-moat/</link><guid isPermaLink="false">https://jacquesthibodeau.com/when-execution-gets-cheap-does-taste-become-the-moat/#audio</guid><description><![CDATA[For twenty years the mantra was that ideas are worthless and execution is everything. That stops holding when execution is cheap, and taste as curation is not what replaces it.]]></description><itunes:summary><![CDATA[For twenty years the mantra was that ideas are worthless and execution is everything. That stops holding when execution is cheap, and taste as curation is not what replaces it.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/69962f8de6be84e878484cc3/audio.mp3" length="7318969" type="audio/mpeg"/><itunes:duration>7:37</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Wed, 18 Feb 2026 00:00:00 GMT</pubDate></item><item><title><![CDATA[Hard Truths About Where AI Is Headed]]></title><link>https://jacquesthibodeau.com/hard-truths-about-where-ai-is-headed/</link><guid isPermaLink="false">https://jacquesthibodeau.com/hard-truths-about-where-ai-is-headed/#audio</guid><description><![CDATA[Four LinkedIn posts collected in one place: coding jobs going away, planning as the scarce skill, rogue models arriving, and what the Pentagon story says about red lines.]]></description><itunes:summary><![CDATA[Four LinkedIn posts collected in one place: coding jobs going away, planning as the scarce skill, rogue models arriving, and what the Pentagon story says about red lines.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/699397f5aff9a9e23424229a/audio.mp3" length="7381663" type="audio/mpeg"/><itunes:duration>7:41</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Mon, 16 Feb 2026 00:00:00 GMT</pubDate></item><item><title><![CDATA[Better model diffing is needed]]></title><link>https://jacquesthibodeau.com/better-model-diffing-is-needed/</link><guid isPermaLink="false">https://jacquesthibodeau.com/better-model-diffing-is-needed/#audio</guid><description><![CDATA[The alignment technique I wish existed. Cheap sensors that show what changed inside a network during training, so training can be steered while it runs rather than audited afterwards.]]></description><itunes:summary><![CDATA[The alignment technique I wish existed. Cheap sensors that show what changed inside a network during training, so training can be steered while it runs rather than audited afterwards.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/6961698f5c58d327551e49e4/audio.mp3" length="3550607" type="audio/mpeg"/><itunes:duration>3:42</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Fri, 09 Jan 2026 00:00:00 GMT</pubDate></item><item><title><![CDATA[Difficulties in Building an AI Safety Startup]]></title><link>https://jacquesthibodeau.com/difficulties-in-building-an-ai-safety-startup/</link><guid isPermaLink="false">https://jacquesthibodeau.com/difficulties-in-building-an-ai-safety-startup/#audio</guid><description><![CDATA[Guardrails, evals, RL environments, control infrastructure. I work through each one and keep landing in the same place, too far from what happens inside a frontier lab.]]></description><itunes:summary><![CDATA[Guardrails, evals, RL environments, control infrastructure. I work through each one and keep landing in the same place, too far from what happens inside a frontier lab.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/69616b115c58d327551e4a41/audio.mp3" length="5401374" type="audio/mpeg"/><itunes:duration>5:38</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Fri, 09 Jan 2026 00:00:00 GMT</pubDate></item><item><title><![CDATA[Gaining clarity on Automated Alignment Research]]></title><link>https://jacquesthibodeau.com/gaining-clarity-on-automated-alignment-research/</link><guid isPermaLink="false">https://jacquesthibodeau.com/gaining-clarity-on-automated-alignment-research/#audio</guid><description><![CDATA[Four different things get conflated in the phrase automated alignment research, and the crux is whether applying techniques we already have is a short step from inventing a new paradigm.]]></description><itunes:summary><![CDATA[Four different things get conflated in the phrase automated alignment research, and the crux is whether applying techniques we already have is a short step from inventing a new paradigm.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/696169f55c58d327551e4a17/audio.mp3" length="6068019" type="audio/mpeg"/><itunes:duration>6:19</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Fri, 09 Jan 2026 00:00:00 GMT</pubDate></item><item><title><![CDATA[Automating AI Safety: What we can do today]]></title><link>https://jacquesthibodeau.com/automating-ai-safety-what-we-can-do-today/</link><guid isPermaLink="false">https://jacquesthibodeau.com/automating-ai-safety-what-we-can-do-today/#audio</guid><description><![CDATA[A concrete list of things to build now so a coding agent can actually run an interpretability experiment, and why the METR slowdown result does not say what people think.]]></description><itunes:summary><![CDATA[A concrete list of things to build now so a coding agent can actually run an interpretability experiment, and why the METR slowdown result does not say what people think.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/68956e635c58d327551e4960/audio.mp3" length="18777870" type="audio/mpeg"/><itunes:duration>19:34</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Fri, 08 Aug 2025 00:00:00 GMT</pubDate></item><item><title><![CDATA[AI Alignment Project Ideas]]></title><link>https://jacquesthibodeau.com/ai-alignment-project-ideas-oct-2-2024/</link><guid isPermaLink="false">https://jacquesthibodeau.com/ai-alignment-project-ideas-oct-2-2024/#audio</guid><description><![CDATA[Seven project write-ups for ARENA and LASR participants, from improving MAIA to reward misspecification and the cases where a model gets the right answer from bad reasoning.]]></description><itunes:summary><![CDATA[Seven project write-ups for ARENA and LASR participants, from improving MAIA to reward misspecification and the cases where a model gets the right answer from bad reasoning.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/66fd6f485c58d327551e4625/audio.mp3" length="12426893" type="audio/mpeg"/><itunes:duration>12:57</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Wed, 02 Oct 2024 00:00:00 GMT</pubDate></item><item><title><![CDATA[How much I'm paying for AI productivity software (and the future of AI use)]]></title><link>https://jacquesthibodeau.com/how-much-im-paying-for-ai-productivity-software/</link><guid isPermaLink="false">https://jacquesthibodeau.com/how-much-im-paying-for-ai-productivity-software/#audio</guid><description><![CDATA[What I was paying for AI tools in September 2024, roughly nine hundred dollars a month, and the harder question underneath it: why it is difficult to spend meaningfully more.]]></description><itunes:summary><![CDATA[What I was paying for AI tools in September 2024, roughly nine hundred dollars a month, and the harder question underneath it: why it is difficult to spend meaningfully more.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/66f2be7d5c58d327551e44b7/audio.mp3" length="13567128" type="audio/mpeg"/><itunes:duration>14:08</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Tue, 24 Sep 2024 00:00:00 GMT</pubDate></item><item><title><![CDATA[The importance of Entropy]]></title><link>https://jacquesthibodeau.com/the-importance-of-entropy/</link><guid isPermaLink="false">https://jacquesthibodeau.com/the-importance-of-entropy/#audio</guid><description><![CDATA[A short physics note. A constant stream of low-entropy energy from the Sun favours structures that dissipate it, which may make life less an accident than a consequence.]]></description><itunes:summary><![CDATA[A short physics note. A constant stream of low-entropy energy from the Sun favours structures that dissipate it, which may make life less an accident than a consequence.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/66bce20e0a5496060d3a1393/audio.mp3" length="3275172" type="audio/mpeg"/><itunes:duration>3:25</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Wed, 14 Aug 2024 00:00:00 GMT</pubDate></item><item><title><![CDATA[Accelerating AI Alignment Research (Talk)]]></title><link>https://jacquesthibodeau.com/accelerating-ai-alignment-research-talk/</link><guid isPermaLink="false">https://jacquesthibodeau.com/accelerating-ai-alignment-research-talk/#audio</guid><description><![CDATA[The keynote I gave to open the research augmentation hackathon I co-organised with Apart Research. Slides and project suggestions are linked underneath.]]></description><itunes:summary><![CDATA[The keynote I gave to open the research augmentation hackathon I co-organised with Apart Research. Slides and project suggestions are linked underneath.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/66a523890a5496060d3a1355/audio.mp3" length="667942" type="audio/mpeg"/><itunes:duration>0:42</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Sat, 27 Jul 2024 00:00:00 GMT</pubDate></item><item><title><![CDATA[Using data attribution for AI alignment]]></title><link>https://jacquesthibodeau.com/using-data-attribution-for-ai-alignment/</link><guid isPermaLink="false">https://jacquesthibodeau.com/using-data-attribution-for-ai-alignment/#audio</guid><description><![CDATA[In-Run Data Shapley is cheap enough to attribute behaviour during pre-training. What I would do with that: order the curriculum so a model holds human values before situational awareness.]]></description><itunes:summary><![CDATA[In-Run Data Shapley is cheap enough to attribute behaviour during pre-training. What I would do with that: order the curriculum so a model holds human values before situational awareness.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/668d67980a5496060d3a1328/audio.mp3" length="14783721" type="audio/mpeg"/><itunes:duration>15:24</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Tue, 09 Jul 2024 00:00:00 GMT</pubDate></item><item><title><![CDATA[Quantum Computing, Photonics, and Energy Bottlenecks for AGI]]></title><link>https://jacquesthibodeau.com/will-quantum-computing-play-a-role-in-agi-development/</link><guid isPermaLink="false">https://jacquesthibodeau.com/will-quantum-computing-play-a-role-in-agi-development/#audio</guid><description><![CDATA[My answer is that quantum probably does not matter and photonics might. The energy numbers, why optical transistors are the wrong thing to judge photonics by, and OpenAI's hire.]]></description><itunes:summary><![CDATA[My answer is that quantum probably does not matter and photonics might. The energy numbers, why optical transistors are the wrong thing to judge photonics by, and OpenAI's hire.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/663e34693de6b56ce4c0cd0a/audio.mp3" length="25191454" type="audio/mpeg"/><itunes:duration>26:14</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Fri, 10 May 2024 00:00:00 GMT</pubDate></item><item><title><![CDATA[AI Insights #1: How Misalignment Could Lead to Takeover & Necessary Safety Properties]]></title><link>https://jacquesthibodeau.com/ai-insights-1-how-misalignment-could-lead-to-takeover-necessary-safety-properties/</link><guid isPermaLink="false">https://jacquesthibodeau.com/ai-insights-1-how-misalignment-could-lead-to-takeover-necessary-safety-properties/#audio</guid><description><![CDATA[A round-up of three: Christiano on takeover thresholds, Shane Legg's necessary properties for any AGI safety plan, and why the danger sits in the agents built on top of models.]]></description><itunes:summary><![CDATA[A round-up of three: Christiano on takeover thresholds, Shane Legg's necessary properties for any AGI safety plan, and why the danger sits in the agents built on top of models.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/6634ddb83de6b56ce4c0cc57/audio.mp3" length="7519591" type="audio/mpeg"/><itunes:duration>7:50</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Fri, 03 May 2024 00:00:00 GMT</pubDate></item><item><title><![CDATA[My current research and request for collaborators]]></title><link>https://jacquesthibodeau.com/update-on-what-ive-been-doing/</link><guid isPermaLink="false">https://jacquesthibodeau.com/update-on-what-ive-been-doing/#audio</guid><description><![CDATA[The bio I wrote for EAG Bay Area 2024: what I was working on, what I wanted to argue about, and the three kinds of collaborator I was looking for.]]></description><itunes:summary><![CDATA[The bio I wrote for EAG Bay Area 2024: what I was working on, what I wanted to argue about, and the three kinds of collaborator I was looking for.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/65b01fd53de6b56ce4c0cbf5/audio.mp3" length="15455800" type="audio/mpeg"/><itunes:duration>16:06</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Tue, 23 Jan 2024 00:00:00 GMT</pubDate></item><item><title><![CDATA[But is it really in Rome? Limitations of the ROME model editing technique]]></title><link>https://jacquesthibodeau.com/but-is-it-really-in-rome-limitations-of-the-rome-model-editing-technique/</link><guid isPermaLink="false">https://jacquesthibodeau.com/but-is-it-really-in-rome-limitations-of-the-rome-model-editing-technique/#audio</guid><description><![CDATA[The ROME edit does not generalise the way you would expect. It runs in one direction only, and cheese and fromage have to be edited separately.]]></description><itunes:summary><![CDATA[The ROME edit does not generalise the way you would expect. It runs in one direction only, and cheese and fromage have to be edited separately.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/63ae4ffa3de6b56ce4c0c922/audio.mp3" length="3238391" type="audio/mpeg"/><itunes:duration>3:22</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Fri, 30 Dec 2022 00:00:00 GMT</pubDate></item><item><title><![CDATA[An incomplete list of projects I'd like to work on in 2023]]></title><link>https://jacquesthibodeau.com/untitled-2/</link><guid isPermaLink="false">https://jacquesthibodeau.com/untitled-2/#audio</guid><description><![CDATA[Two sentences pointing at a LessWrong shortform where I listed the projects I wanted to work on in 2023.]]></description><itunes:summary><![CDATA[Two sentences pointing at a LessWrong shortform where I listed the projects I wanted to work on in 2023.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/63adbb913de6b56ce4c0c910/audio.mp3" length="183946" type="audio/mpeg"/><itunes:duration>0:11</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Thu, 29 Dec 2022 00:00:00 GMT</pubDate></item><item><title><![CDATA[(Linkpost) Results for a survey of tool use and workflows in alignment research]]></title><link>https://jacquesthibodeau.com/linkpost/</link><guid isPermaLink="false">https://jacquesthibodeau.com/linkpost/#audio</guid><description><![CDATA[A pointer to the results of the March 2022 survey on tool use in alignment research. The write-up itself is on LessWrong.]]></description><itunes:summary><![CDATA[A pointer to the results of the March 2022 survey on tool use in alignment research. The write-up itself is on LessWrong.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/63a086223de6b56ce4c0c8f2/audio.mp3" length="523328" type="audio/mpeg"/><itunes:duration>0:33</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Mon, 19 Dec 2022 00:00:00 GMT</pubDate></item><item><title><![CDATA[How learning efficiently applies to alignment research]]></title><link>https://jacquesthibodeau.com/regarding-how-learning-efficiently-applies-to-alignment-research/</link><guid isPermaLink="false">https://jacquesthibodeau.com/regarding-how-learning-efficiently-applies-to-alignment-research/#audio</guid><description><![CDATA[Studying only counts here if it shortens the path to insights that solve alignment. Wentworth's point about newcomers losing years to low-value sub-problems is the thing to avoid.]]></description><itunes:summary><![CDATA[Studying only counts here if it shortens the path to insights that solve alignment. Wentworth's point about newcomers losing years to low-value sub-problems is the thing to avoid.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/639cd0913de6b56ce4c0c8df/audio.mp3" length="2354826" type="audio/mpeg"/><itunes:duration>2:27</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Fri, 16 Dec 2022 00:00:00 GMT</pubDate></item><item><title><![CDATA[Differential Training Process: Delaying capabilities until inner aligned]]></title><link>https://jacquesthibodeau.com/differential-training-process/</link><guid isPermaLink="false">https://jacquesthibodeau.com/differential-training-process/#audio</guid><description><![CDATA[Keep the model from working out that it is in a training loop until its values are where we want them. Plus the two objections I have not answered.]]></description><itunes:summary><![CDATA[Keep the model from working out that it is in a training loop until its values are where we want them. Plus the two objections I have not answered.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/6390be413de6b56ce4c0c8c5/audio.mp3" length="3994523" type="audio/mpeg"/><itunes:duration>4:10</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Wed, 07 Dec 2022 00:00:00 GMT</pubDate></item><item><title><![CDATA[Near-Term AI capabilities probably bring low-hanging fruits for global poverty/health]]></title><link>https://jacquesthibodeau.com/untitled/</link><guid isPermaLink="false">https://jacquesthibodeau.com/untitled/#audio</guid><description><![CDATA[A call to charity entrepreneurs in December 2022. Models of that generation were about to make some interventions viable, and the non-profit sector had not caught up.]]></description><itunes:summary><![CDATA[A call to charity entrepreneurs in December 2022. Models of that generation were about to make some interventions viable, and the non-profit sector had not caught up.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/6390bcd73de6b56ce4c0c8a5/audio.mp3" length="1235113" type="audio/mpeg"/><itunes:duration>1:17</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Wed, 07 Dec 2022 00:00:00 GMT</pubDate></item><item><title><![CDATA[Foresight for AGI Safety Strategy]]></title><link>https://jacquesthibodeau.com/foresight-for-agi-safety-strategy/</link><guid isPermaLink="false">https://jacquesthibodeau.com/foresight-for-agi-safety-strategy/#audio</guid><description><![CDATA[Governance people should be writing the policy document and building the demo for each plausible scenario now, rather than starting the day a minister gives them three months.]]></description><itunes:summary><![CDATA[Governance people should be writing the policy document and building the demo for each plausible scenario now, rather than starting the day a minister gives them three months.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/638e25343de6b56ce4c0c880/audio.mp3" length="23232017" type="audio/mpeg"/><itunes:duration>24:12</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Mon, 05 Dec 2022 00:00:00 GMT</pubDate></item><item><title><![CDATA[Is the "Valley of Confused Abstractions" real?]]></title><link>https://jacquesthibodeau.com/is-the-valley-of-confused-abstractions-real/</link><guid isPermaLink="false">https://jacquesthibodeau.com/is-the-valley-of-confused-abstractions-real/#audio</guid><description><![CDATA[Chris Olah's curve says models get harder to read before they get easier. Neel told me that came out of vision models, so I am posting my confusion.]]></description><itunes:summary><![CDATA[Chris Olah's curve says models get harder to read before they get easier. Neel told me that came out of vision models, so I am posting my confusion.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/638e25c83de6b56ce4c0c891/audio.mp3" length="3963594" type="audio/mpeg"/><itunes:duration>4:08</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Mon, 05 Dec 2022 00:00:00 GMT</pubDate></item><item><title><![CDATA[Notes on Cicero]]></title><link>https://jacquesthibodeau.com/notes-on-cicero/</link><guid isPermaLink="false">https://jacquesthibodeau.com/notes-on-cicero/#audio</guid><description><![CDATA[Cicero never lies. It states the plan it holds and then changes plan, and that is what players experienced as betrayal. Interpretability on the model alone would have missed it.]]></description><itunes:summary><![CDATA[Cicero never lies. It states the plan it holds and then changes plan, and that is what players experienced as betrayal. Interpretability on the model alone would have missed it.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/6384d5a33de6b56ce4c0c82c/audio.mp3" length="4093162" type="audio/mpeg"/><itunes:duration>4:16</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Mon, 28 Nov 2022 00:00:00 GMT</pubDate></item><item><title><![CDATA[Detail about factual knowledge in Transformers]]></title><link>https://jacquesthibodeau.com/detail-about-the-rome-paper-ive-been-thinking-about/</link><guid isPermaLink="false">https://jacquesthibodeau.com/detail-about-the-rome-paper-ive-been-thinking-about/#audio</guid><description><![CDATA[The model writes a spread of facts about the subject into the residual stream before it knows what will be asked. An appendix to the ROME post.]]></description><itunes:summary><![CDATA[The model writes a spread of facts about the subject into the residual stream before it knows what will be asked. An appendix to the ROME post.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/638229233de6b56ce4c0c7f5/audio.mp3" length="2169252" type="audio/mpeg"/><itunes:duration>2:16</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Sat, 26 Nov 2022 00:00:00 GMT</pubDate></item><item><title><![CDATA[Current Thoughts on my Learning System]]></title><link>https://jacquesthibodeau.com/current-thoughts-on-learning-systems/</link><guid isPermaLink="false">https://jacquesthibodeau.com/current-thoughts-on-learning-systems/#audio</guid><description><![CDATA[I coasted on raw ability and it stopped being enough. Learning is a set of skills you have to practise, and this is my public commitment to build the system.]]></description><itunes:summary><![CDATA[I coasted on raw ability and it stopped being enough. Learning is a set of skills you have to practise, and this is my public commitment to build the system.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/62f83dba3bc733057f4585d2/audio.mp3" length="5398448" type="audio/mpeg"/><itunes:duration>5:37</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Sun, 14 Aug 2022 00:00:00 GMT</pubDate></item><item><title><![CDATA[What does "Effective" in EA mean to you?]]></title><link>https://jacquesthibodeau.com/what-does-effective-in-ea-mean-to-you/</link><guid isPermaLink="false">https://jacquesthibodeau.com/what-does-effective-in-ea-mean-to-you/#audio</guid><description><![CDATA[Effective, to me, always meant doing the most good inside the share of your life you have decided to give it. A movement too demanding of everyone collapses.]]></description><itunes:summary><![CDATA[Effective, to me, always meant doing the most good inside the share of your life you have decided to give it. A movement too demanding of everyone collapses.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/62e0d2473bc733057f4585b4/audio.mp3" length="3418532" type="audio/mpeg"/><itunes:duration>3:34</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Wed, 27 Jul 2022 00:00:00 GMT</pubDate></item><item><title><![CDATA[Helping organizations survive disasters (and potentially avoid them altogether)]]></title><link>https://jacquesthibodeau.com/helping-organizations-survive-disasters-and-potentially-avoid-them-altogether/</link><guid isPermaLink="false">https://jacquesthibodeau.com/helping-organizations-survive-disasters-and-potentially-avoid-them-altogether/#audio</guid><description><![CDATA[My 2019 primer on strategic foresight, reshared for AGI forecasters. Low-probability high-impact developments get ignored because they are unlikely, and Bostrom's urn is the reason to look anyway.]]></description><itunes:summary><![CDATA[My 2019 primer on strategic foresight, reshared for AGI forecasters. Low-probability high-impact developments get ignored because they are unlikely, and Bostrom's urn is the reason to look anyway.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/62dc37093bc733057f458550/audio.mp3" length="11809986" type="audio/mpeg"/><itunes:duration>12:18</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Sat, 23 Jul 2022 00:00:00 GMT</pubDate></item><item><title><![CDATA[AI Alignment YouTube Playlists]]></title><link>https://jacquesthibodeau.com/ai-alignment-youtube-playlists/</link><guid isPermaLink="false">https://jacquesthibodeau.com/ai-alignment-youtube-playlists/#audio</guid><description><![CDATA[Two YouTube playlists of alignment talks, split by whether you have to watch the slides, so that one of them works while you walk.]]></description><itunes:summary><![CDATA[Two YouTube playlists of alignment talks, split by whether you have to watch the slides, so that one of them works while you walk.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/62da3c583bc733057f458506/audio.mp3" length="889461" type="audio/mpeg"/><itunes:duration>0:56</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Fri, 22 Jul 2022 00:00:00 GMT</pubDate></item><item><title><![CDATA[I'll be in Berkeley for SERI MATS for the next 2 months]]></title><link>https://jacquesthibodeau.com/ill-be-in-berkeley-for-seri-mats/</link><guid isPermaLink="false">https://jacquesthibodeau.com/ill-be-in-berkeley-for-seri-mats/#audio</guid><description><![CDATA[A short note from July 2022 saying I was heading to Berkeley for SERI MATS, and that I would like to meet people while I was there.]]></description><itunes:summary><![CDATA[A short note from July 2022 saying I was heading to Berkeley for SERI MATS, and that I would like to meet people while I was there.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/62daddfe3bc733057f458531/audio.mp3" length="708902" type="audio/mpeg"/><itunes:duration>0:44</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Fri, 22 Jul 2022 00:00:00 GMT</pubDate></item><item><title><![CDATA[A descriptive, not prescriptive, overview of current AI Alignment Research]]></title><link>https://jacquesthibodeau.com/a-descriptive-not-prescriptive-overview-of-current-ai-alignment-research/</link><guid isPermaLink="false">https://jacquesthibodeau.com/a-descriptive-not-prescriptive-overview-of-current-ai-alignment-research/#audio</guid><description><![CDATA[We catalogued the alignment literature and let the analysis say which research directions actually exist. The AI Safety Camp project behind the arXiv paper and the alignment research dataset.]]></description><itunes:summary><![CDATA[We catalogued the alignment literature and let the analysis say which research directions actually exist. The AI Safety Camp project behind the arXiv paper and the alignment research dataset.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/62d9c381a7117a1815e9c590/audio.mp3" length="2233199" type="audio/mpeg"/><itunes:duration>2:20</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Thu, 21 Jul 2022 00:00:00 GMT</pubDate></item><item><title><![CDATA[A survey of tool use and workflows in alignment research]]></title><link>https://jacquesthibodeau.com/a-survey-of-tool-use-and-workflows-in-alignment-research/</link><guid isPermaLink="false">https://jacquesthibodeau.com/a-survey-of-tool-use-and-workflows-in-alignment-research/#audio</guid><description><![CDATA[Before building anything we asked alignment researchers how they work and where a language model tool would earn its place. The survey is closed, the dual-use note still stands.]]></description><itunes:summary><![CDATA[Before building anything we asked alignment researchers how they work and where a language model tool would earn its place. The survey is closed, the dual-use note still stands.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/62d9c28aa7117a1815e9c568/audio.mp3" length="2489408" type="audio/mpeg"/><itunes:duration>2:36</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Wed, 23 Mar 2022 00:00:00 GMT</pubDate></item><item><title><![CDATA[Interesting Applications of GPT-3: Elicit]]></title><link>https://jacquesthibodeau.com/interesting-applications-of-gpt-3-elicit/</link><guid isPermaLink="false">https://jacquesthibodeau.com/interesting-applications-of-gpt-3-elicit/#audio</guid><description><![CDATA[My first post on this site, walking through Ought's Elicit in 2021. The subquestion decomposition is the earliest form of the idea I have been working on ever since.]]></description><itunes:summary><![CDATA[My first post on this site, walking through Ought's Elicit in 2021. The subquestion decomposition is the earliest form of the idea I have been working on ever since.]]></itunes:summary><enclosure url="https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/62d9947aa7117a1815e9c4dc/audio.mp3" length="13639854" type="audio/mpeg"/><itunes:duration>14:12</itunes:duration><itunes:author><![CDATA[Jacques Thibodeau]]></itunes:author><itunes:explicit>false</itunes:explicit><pubDate>Fri, 24 Sep 2021 00:00:00 GMT</pubDate></item></channel></rss>