The Fractured Entangled Representation Hypothesis (Kenneth Stanley, Akarsh Kumar)

<p>Are the AI models you use today imposters?</p><p><br></p><p>Please watch the intro video we did before this: https://www.youtube.com/watch?v=o1q6Hhz0MAg</p><p><br></p><p>In this episode, hosts Dr. Tim Scarfe and Dr. Duggar are joined by AI researcher Prof. Kenneth Stanley and MIT PhD student Akash Kumar to discuss their fascinating paper, &quot;Questioning Representational Optimism in Deep Learning.&quot;</p><p><br></p><p>Imagine you ask two people to draw a perfect skull. One is a brilliant artist who understands anatomy, the other is a machine that just traces the image. Both drawings look identical, but the artist understands what a skull is—they know where the mouth is, how the jaw works, and that it&#39;s symmetrical. The machine just has a tangled mess of lines that happens to form the right picture.</p><p><br></p><p>An AI with an elegant representation, has the building blocks to generate truly new ideas.</p><p><br></p><p>The Path Is the Goal: As Kenneth Stanley puts it, &quot;it matters not just where you get, but how you got there&quot;. Two students can ace a math test, but the one who truly understands the concepts—instead of just memorizing formulas—is the one who will go on to make new discoveries.</p><p><br></p><p>The show is a mixture of 3 separate recordings we have done, the original Patreon warmup with Tim/Kenneth, the Tim/Keith &quot;Steakhouse&quot; recorded after the main interview, then the main interview with Kenneth/Akarsh/Keith/Tim. Feel free to skip around. We had to edit this in a rush as we are travelling next week but it&#39;s reasonably cleaned up. </p><p><br></p><p>TOC:</p><p>00:00:00 Intro: Garbage vs. Amazing Representations</p><p>00:05:42 How Good Representations Form</p><p>00:11:14 Challenging the &quot;Bitter Lesson&quot;</p><p>00:18:04 AI Creativity &amp; Representation Types</p><p>00:22:13 Steakhouse: Critiques &amp; Alternatives</p><p>00:28:30 Steakhouse: Key Concepts &amp; Goldilocks Zone</p><p>00:39:42 Steakhouse: A Sober View on AI Risk</p><p>00:43:46 Steakhouse: The Paradox of Open-Ended Search</p><p>00:47:58 Main Interview: Paper Intro &amp; Core Concepts</p><p>00:56:44 Main Interview: Deception and Evolvability</p><p>01:36:30 Main Interview: Reinterpreting Evolution</p><p>01:56:16 Main Interview: Impostor Intelligence</p><p>02:11:15 Main Interview: Recommendations for AI Research</p><p><br></p><p>REFS:</p><p>Questioning Representational Optimism in Deep Learning:</p><p>The Fractured Entangled Representation Hypothesis</p><p>Akarsh Kumar, Jeff Clune, Joel Lehman, Kenneth O. Stanley</p><p>https://arxiv.org/pdf/2505.11581</p><p><br></p><p>Kenneth O. Stanley, Joel Lehman</p><p>Why Greatness Cannot Be Planned: The Myth of the Objective</p><p>https://amzn.to/44xLaXK</p><p><br></p><p>Original show with Kenneth from 4 years ago:</p><p>https://www.youtube.com/watch?v=lhYGXYeMq_E</p><p><br></p><p>Kenneth Stanley is SVP Open Endedness at Lila Sciences</p><p>https://x.com/kenneth0stanley</p><p><br></p><p>Akarsh Kumar (MIT)</p><p>https://akarshkumar.com/</p><p><br></p><p>AND... Kenneth is HIRING (this is an OPPORTUNITY OF A LIFETIME!)</p><p>Research Engineer: https://job-boards.greenhouse.io/lila/jobs/7890007002</p><p>Research Scientist: https://job-boards.greenhouse.io/lila/jobs/8012245002</p><p><br></p><p>TRANSCRIPT:</p><p>https://app.rescript.info/public/share/W_T7E1OC2Wj49ccqlIOOztg2MJWaaVbovTeyxcFEQdU</p>