Recent Podcast Interview with Demis Hassabis

Source:https://chatgpt.com/share/67a0ad1b-c808-8012-811f-2abfca6ca174

I'll search for a recent podcast interview featuring Demis Hassabis within the last six months. The interviewer might be someone with a name similar to 'Davashi'—I'll look for any matches across major podcast platforms like YouTube, Spotify, and Apple Podcasts.

I'll get back to you once I find the details.

Recent Podcast Interview with Demis Hassabis (Hosted by Dwarkesh Patel)

Episode Title, Host, and Date

  • Title: “Demis Hassabis - Scaling, Superhuman AIs, AlphaZero atop LLMs, Rogue Nations Threat.”www.dwarkeshpatel.com
  • Host: Dwarkesh Patel (host of the Dwarkesh Podcast)www.dwarkeshpatel.com– The interviewer’s name “Dwarkesh” may have sounded like “Davashi.”
  • Release Date: February 28, 2024www.dwarkeshpatel.com. (This falls within the last year; it’s a recent interview featuring Demis Hassabis.)

Key Discussion Points

In this one-hour conversation, Demis Hassabis (CEO of Google DeepMind) and host Dwarkesh Patel cover several insightful topics

www.dwarkeshpatel.com:

  • Scaling as an Artform: Hassabis explains why scaling AI models is an artform, emphasizing the challenges and nuances in scaling up AI systems effectivelywww.dwarkeshpatel.com.
  • Combining Search/Planning with LLMs: They discuss adding search, planning, and AlphaZero-type training atop large language models (LLMs) – essentially how techniques from DeepMind’s AlphaZero can enhance LLM capabilitieswww.dwarkeshpatel.com.
  • Securing AI Weights: The conversation addresses how to ensure “rogue nations” or bad actors can’t steal AI model weights, highlighting the importance of security and safe deployment of advanced AIwww.dwarkeshpatel.com.
  • Aligning Superhuman AI: Hassabis talks about the right way to align superhuman AIs and manage an intelligence explosion, i.e. how to safely guide AI if it rapidly becomes far more intelligent (often termed the AI alignment problem)www.dwarkeshpatel.com.

These points provide a glimpse into cutting-edge AI issues like scaling strategies, integrating different AI approaches, security concerns, and AI alignment.

Where to Listen/Watch

This interview is available on multiple major platforms

www.dwarkeshpatel.com:

Each of these links provides access to the full interview, titled “Demis Hassabis - Scaling, Superhuman AIs, AlphaZero atop LLMs, Rogue Nations Threat,” with host Dwarkesh Patel (the likely “Davashi”-sounding name)

www.dwarkeshpatel.comwww.dwarkeshpatel.com. The discussion offers valuable insights into Demis Hassabis’s views on the future of AI, making it a must-listen for those interested in artificial intelligence and DeepMind’s work.

Sources: The episode details and summary are confirmed by Dwarkesh Patel’s official podcast notes

www.dwarkeshpatel.comwww.dwarkeshpatel.comand the Apple Podcasts listingpodcasts.apple.compodcasts.apple.comfor this interview. The podcast was released on Feb 28, 2024www.dwarkeshpatel.com.