Paul Christiano - Preventing an AI Takeover

<p>Paul Christiano is the world’s leading AI safety researcher. My full episode with him is out!</p><p>We discuss:</p><p>- Does he regret inventing RLHF, and is alignment necessarily dual-use?</p><p>- Why he has relatively modest timelines (40% by 2040, 15% by 2030),</p><p>- What do we want post-AGI world to look like (do we want to keep gods enslaved forever)?</p><p>- Why he’s leading the push to get to labs develop responsible scaling policies, and what it would take to prevent an AI coup or bioweapon,</p><p>- His current research into a new proof system, and how this could solve alignment by explaining model's behavior</p><p>- and much more.</p><p>Watch on <a target="_blank" href="https://youtu.be/9AAhTLa0dT0">YouTube</a>. Listen on <a target="_blank" href="https://podcasts.apple.com/us/podcast/paul-christiano-preventing-an-ai-takeover/id1516093381?i=1000633226398">Apple Podcasts</a>, <a target="_blank" href="https://open.spotify.com/episode/5vOuxDP246IG4t4K3EuEKj?si=VW7qTs8ZRHuQX9emnboGcA">Spotify</a>, or any other podcast platform. Read the full transcript <a target="_blank" href="https://www.dwarkeshpatel.com/i/138448008/transcript">here</a>. Follow <a target="_blank" href="https://twitter.com/dwarkesh_sp">me on Twitter</a> for updates on future episodes.</p><p></p><p><strong>Open Philanthropy</strong></p><p>Open Philanthropy is currently hiring for <strong>twenty-two</strong> different roles to reduce catastrophic risks from fast-moving advances in AI and biotechnology, including grantmaking, research, and operations.</p><p>For more information and to apply, please see <a target="_blank" href="https://www.openphilanthropy.org/research/new-roles-on-our-gcr-team/">the application</a>: <a target="_blank" href="https://www.openphilanthropy.org/research/new-roles-on-our-gcr-team/">https://www.openphilanthropy.org/research/new-roles-on-our-gcr-team/</a></p><p>The deadline to apply is <strong>November 9th</strong>; make sure to check out those roles before they close.</p><p></p><p><strong>Timestamps</strong></p><