Shane Legg (DeepMind Founder) — 2028 AGI, superhuman alignment, new architectures

ℹ️ EPISODE INFO

Released

October 26, 2023

Duration

44m

Channel

Dwarkesh Podcast

Watch on YouTube

▶ Open ↗

📝 EPISODE DESCRIPTION

I had a lot of fun chatting with Shane Legg - Founder & Chief AGI Scientist, Google DeepMind! We discuss:

• Why he expects AGI around 2028
• How to align superhuman models
• What new architectures needed for AGI
• Has Deepmind sped up capabilities or safety more?
• Why multimodality will be next big landmark
• & much more

𝐄𝐏𝐈𝐒𝐎𝐃𝐄 𝐋𝐈𝐍𝐊𝐒

• Transcript: https://www.dwarkeshpatel.com/p/shane-legg
• Apple Podcasts: https://podcasts.apple.com/us/podcast/shane-legg-deepmind-founder-2028-agi-new-architectures/id1516093381?i=1000632720307
• Spotify: https://open.spotify.com/episode/0Ru2CtaJqsQ5mpA5dqHWAK?si=4AsglwIZQpqht7p9Wpc_CA
• Twitter: https://twitter.com/dwarkesh_sp/status/1717566262472237134

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒 00:00:00 - Measuring AGI 00:11:41 - Do we need new architectures? 00:16:26 - Is search needed for creativity? 00:19:19 - Superhuman alignment 00:29:58 - Impact of Deepmind on safety vs capabilities 00:34:03 - Timelines 00:41:24 - Multimodality

🎤 SPEAKERS

• Dwarkesh Patel

  host

• Shane Legg

  guest

• Narrator

  other

🧾 EPISODE SUMMARY

In this episode of Dwarkesh Podcast, featuring Dwarkesh Patel and Shane Legg, Shane Legg (DeepMind Founder) — 2028 AGI, superhuman alignment, new architectures explores deepMind’s Shane Legg Predicts 2028 AGI, Maps Path and Risks Shane Legg, co‑founder and Chief AGI Scientist at Google DeepMind, discusses how to define and measure AGI as human‑level, broadly general cognitive capability across many domains rather than a single benchmark. He argues current large language models have unlocked a scalable form of understanding but still lack key ingredients like episodic memory, robust system‑2 reasoning, and grounded multimodal perception. Legg is cautiously optimistic that architectural advances, better search/agency on top of foundation models, and improved factuality/memory will remove most remaining technical blockers, making AGI by around 2028 roughly a 50% probability. On safety, he believes containment will fail for very capable systems, so alignment must come from deeply embedding explicit ethical reasoning and oversight into powerful, well‑informed world models, alongside institutional safeguards and evolving safety benchmarks.

🔗 RELATED EPISODES

General relativity from first principles – Adam Brown

Grant Sanderson (@3blue1brown) – AI disproved a famous math conjecture. Now what?

@Asianometry & Dylan Patel — How the semiconductor industry actually works

Ilya Sutskever (OpenAI Chief Scientist) — Why next-token prediction could surpass human intelligence

Daniel Yergin — Oil destroyed Hitler, fracking destroyed Putin

Lars Doucet — Progress, poverty, Georgism, & why rent is too damn high