#416 – Yann Lecun: Meta AI, Open Source, Limits of LLMs, AGI & the Future of AI

O tym odcinku

<p>Yann LeCun is the Chief AI Scientist at Meta, professor at NYU, Turing Award winner, and one of the most influential researchers in the history of AI. Please support this podcast by checking out our sponsors:<br /> &#8211; <b>HiddenLayer</b>: <a href="https://hiddenlayer.com/lex">https://hiddenlayer.com/lex</a><br /> &#8211; <b>LMNT</b>: <a href="https://drinkLMNT.com/lex">https://drinkLMNT.com/lex</a> to get free sample pack<br /> &#8211; <b>Shopify</b>: <a href="https://shopify.com/lex">https://shopify.com/lex</a> to get $1 per month trial<br /> &#8211; <b>AG1</b>: <a href="https://drinkag1.com/lex">https://drinkag1.com/lex</a> to get 1 month supply of fish oil</p> <p>Transcript: <a href="https://lexfridman.com/yann-lecun-3-transcript">https://lexfridman.com/yann-lecun-3-transcript</a></p> <p><b>EPISODE LINKS:</b><br /> Yann&#8217;s Twitter: <a href="https://twitter.com/ylecun">https://twitter.com/ylecun</a><br /> Yann&#8217;s Facebook: <a href="https://facebook.com/yann.lecun">https://facebook.com/yann.lecun</a><br /> Meta AI: <a href="https://ai.meta.com/">https://ai.meta.com/</a></p> <p><b>PODCAST INFO:</b><br /> Podcast website: <a href="https://lexfridman.com/podcast">https://lexfridman.com/podcast</a><br /> Apple Podcasts: <a href="https://apple.co/2lwqZIr">https://apple.co/2lwqZIr</a><br /> Spotify: <a href="https://spoti.fi/2nEwCF8">https://spoti.fi/2nEwCF8</a><br /> RSS: <a href="https://lexfridman.com/feed/podcast/">https://lexfridman.com/feed/podcast/</a><br /> YouTube Full Episodes: <a href="https://youtube.com/lexfridman">https://youtube.com/lexfridman</a><br /> YouTube Clips: <a href="https://youtube.com/lexclips">https://youtube.com/lexclips</a></p> <p><b>SUPPORT &#38; CONNECT:</b><br /> &#8211; Check out the sponsors above, it&#8217;s the best way to support this podcast<br /> &#8211; Support on Patreon: <a href="https://www.patreon.com/lexfridman">https://www.patreon.com/lexfridman</a><br /> &#8211; Twitter: <a href="https://twitter.com/lexfridman">https://twitter.com/lexfridman</a><br /> &#8211; Instagram: <a href="https://www.instagram.com/lexfridman">https://www.instagram.com/lexfridman</a><br /> &#8211; LinkedIn: <a href="https://www.linkedin.com/in/lexfridman">https://www.linkedin.com/in/lexfridman</a><br /> &#8211; Facebook: <a href="https://www.facebook.com/lexfridman">https://www.facebook.com/lexfridman</a><br /> &#8211; Medium: <a href="https://medium.com/@lexfridman">https://medium.com/@lexfridman</a></p> <p><b>OUTLINE:</b><br /> Here&#8217;s the timestamps for the episode. On some podcast players you should be able to click the timestamp to jump to that time.<br /> (00:00) &#8211; Introduction<br /> (09:10) &#8211; Limits of LLMs<br /> (20:47) &#8211; Bilingualism and thinking<br /> (24:39) &#8211; Video prediction<br /> (31:59) &#8211; JEPA (Joint-Embedding Predictive Architecture)<br /> (35:08) &#8211; JEPA vs LLMs<br /> (44:24) &#8211; DINO and I-JEPA<br /> (45:44) &#8211; V-JEPA<br /> (51:15) &#8211; Hierarchical planning<br /> (57:33) &#8211; Autoregressive LLMs<br /> (1:12:59) &#8211; AI hallucination<br /> (1:18:23) &#8211; Reasoning in AI<br /> (1:35:55) &#8211; Reinforcement learning<br /> (1:41:02) &#8211; Woke AI<br /> (1:50:41) &#8211; Open source<br /> (1:54:19) &#8211; AI and ideology<br /> (1:56:50) &#8211; Marc Andreesen<br /> (2:04:49) &#8211; Llama 3<br /> (2:11:13) &#8211; AGI<br /> (2:15:41) &#8211; AI doomers<br /> (2:31:31) &#8211; Joscha Bach<br /> (2:35:44) &#8211; Humanoid robots<br /> (2:44:52) &#8211; Hope for the future</p>