Ep 58: Google Researchers Noam Shazeer and Jack Rae on Scaling Test-time Compute, Reactions to Ilya & AGI

On the latest episode of Unsupervised Learning, Jacob is joined by two of the most influential minds in AI today. 

🔹 Noam Shazeer, co-inventor of the Transformer

🔹 Jack Rae, Research Director at DeepMind and one of the leads behind Gemini’s Flash Thinking

We got to ask them all of the top-of-mind questions in AI today about where we are, where we’re headed and what it means for businesses and the world. Some key take-aways:

(0:00) Intro
(1:30) Exploring Gemini 2.0
(4:04) Challenges with Evals and Benchmarks
(6:14) Reinforcement Loops and AI Productivity
(8:15) Agentic Coding and AI in Development
(13:02) Multimodal Capabilities and Applications
(15:21) Future of AI: Complexity and Reliability
(19:02) Test Time Compute and Data Efficiency
(31:20) AI Research Culture and Breakthroughs
(38:59) Reflecting on Large Language Models
(39:37) The Rise of Vision-Based Models
(41:01) Native Image Generation and General Purpose Models
(41:35) AI in Healthcare and Specialized Models
(43:32) Shifting Timelines and Rapid Adoption
(46:48) Open Source Models and Competitive Edge
(49:30) AI's Impact on Education and Personal Lives
(55:10) AGI Risks and Safety Considerations
(57:33) Future of AI Companions
(1:02:17) Quickfire

With your co-hosts: 

@jacobeffron 

- Partner at Redpoint, Former PM Flatiron Health 

@patrickachase 

- Partner at Redpoint, Former ML Engineer LinkedIn 

@ericabrescia 

- Former COO Github, Founder Bitnami (acq’d by VMWare) 

@jordan_segall 

- Partner at Redpoint