<description>&lt;p&gt;The &lt;a target="_blank" href="https://x.com/karpathy"&gt;Andrej Karpathy&lt;/a&gt; episode.&lt;/p&gt;&lt;p&gt;During this interview, Andrej explains why reinforcement learning is terrible (but everything else is much worse), why AGI will just blend into the previous ~2.5 centuries of 2% GDP growth, why self driving took so long to crack, and what he sees as the future of education.&lt;/p&gt;&lt;p&gt;It was a pleasure chatting with him.&lt;/p&gt;&lt;p&gt;Watch on &lt;a target="_blank" href="https://youtu.be/lXUZvyajciY"&gt;YouTube&lt;/a&gt;; read the &lt;a target="_blank" href="https://dwarkesh.substack.com/p/andrej-karpathy"&gt;transcript&lt;/a&gt;.&lt;/p&gt;&lt;p&gt;Sponsors&lt;/p&gt;&lt;p&gt;* &lt;a target="_blank" href="https://labelbox.com/dwarkesh"&gt;Labelbox&lt;/a&gt; helps you get data that is more detailed, more accurate, and higher signal than you could get by default, no matter your domain or training paradigm. Reach out today at &lt;a target="_blank" href="https://labelbox.com/dwarkesh"&gt;labelbox.com/dwarkesh&lt;/a&gt;&lt;/p&gt;&lt;p&gt;* &lt;a target="_blank" href="https://mercury.com"&gt;Mercury&lt;/a&gt; helps you run your business better. It’s the banking platform we use for the podcast — we love that we can see our accounts, cash flows, AR, and AP all in one place. Apply online in minutes at &lt;a target="_blank" href="https://mercury.com"&gt;mercury.com&lt;/a&gt;&lt;/p&gt;&lt;p&gt;* &lt;a target="_blank" href="https://blog.google/technology/ai/veo-updates-flow/"&gt;Google’s Veo 3.1&lt;/a&gt; update is a notable improvement to an already great model. Veo 3.1’s generations are more coherent and the audio is even higher-quality. If you have a Google AI Pro or Ultra plan, you can try it in Gemini today by visiting &lt;a target="_blank" href="https://gemini.google"&gt;https://gemini.google&lt;/a&gt;&lt;/p&gt;&lt;p&gt;Timestamps&lt;/p&gt;&lt;p&gt;(00:00:00) – AGI is still a decade away&lt;/p&gt;&lt;p&gt;(00:29:45) – LLM cognitive deficits&lt;/p&gt;&lt;p&gt;(00:40:05) – RL is terrible&lt;/p&gt;&lt;p&gt;(00:49:38) – How do humans learn?&lt;/p&gt;&lt;p&gt;(01:06:25) – AGI will blend into 2% GDP growth&lt;/p&gt;&lt;p&gt;(01:17:36) – ASI&lt;/p&gt;&lt;p&gt;(01:32:50) – Evolution of intelligence &amp; culture&lt;/p&gt;&lt;p&gt;(01:42:55) - Why self driving took so long&lt;/p&gt;&lt;p&gt;(01:56:20) - Future of education&lt;/p&gt; &lt;br/&gt;&lt;br/&gt;Get full access to Dwarkesh Podcast at &lt;a href="https://www.dwarkesh.com/subscribe?utm_medium=podcast&amp;#38;utm_campaign=CTA_4"&gt;www.dwarkesh.com/subscribe&lt;/a&gt;</description>

Dwarkesh Podcast

Dwarkesh Patel

Andrej Karpathy — AGI is still a decade away

OCT 17, 2025145 MIN
Dwarkesh Podcast

Andrej Karpathy — AGI is still a decade away

OCT 17, 2025145 MIN

Description

<p>The <a target="_blank" href="https://x.com/karpathy">Andrej Karpathy</a> episode.</p><p>During this interview, Andrej explains why reinforcement learning is terrible (but everything else is much worse), why AGI will just blend into the previous ~2.5 centuries of 2% GDP growth, why self driving took so long to crack, and what he sees as the future of education.</p><p>It was a pleasure chatting with him.</p><p>Watch on <a target="_blank" href="https://youtu.be/lXUZvyajciY">YouTube</a>; read the <a target="_blank" href="https://dwarkesh.substack.com/p/andrej-karpathy">transcript</a>.</p><p>Sponsors</p><p>* <a target="_blank" href="https://labelbox.com/dwarkesh">Labelbox</a> helps you get data that is more detailed, more accurate, and higher signal than you could get by default, no matter your domain or training paradigm. Reach out today at <a target="_blank" href="https://labelbox.com/dwarkesh">labelbox.com/dwarkesh</a></p><p>* <a target="_blank" href="https://mercury.com">Mercury</a> helps you run your business better. It’s the banking platform we use for the podcast — we love that we can see our accounts, cash flows, AR, and AP all in one place. Apply online in minutes at <a target="_blank" href="https://mercury.com">mercury.com</a></p><p>* <a target="_blank" href="https://blog.google/technology/ai/veo-updates-flow/">Google’s Veo 3.1</a> update is a notable improvement to an already great model. Veo 3.1’s generations are more coherent and the audio is even higher-quality. If you have a Google AI Pro or Ultra plan, you can try it in Gemini today by visiting <a target="_blank" href="https://gemini.google">https://gemini.google</a></p><p>Timestamps</p><p>(00:00:00) – AGI is still a decade away</p><p>(00:29:45) – LLM cognitive deficits</p><p>(00:40:05) – RL is terrible</p><p>(00:49:38) – How do humans learn?</p><p>(01:06:25) – AGI will blend into 2% GDP growth</p><p>(01:17:36) – ASI</p><p>(01:32:50) – Evolution of intelligence & culture</p><p>(01:42:55) - Why self driving took so long</p><p>(01:56:20) - Future of education</p> <br/><br/>Get full access to Dwarkesh Podcast at <a href="https://www.dwarkesh.com/subscribe?utm_medium=podcast&#38;utm_campaign=CTA_4">www.dwarkesh.com/subscribe</a>