The End of GPU Scaling? Compute & The Agent Era — Tim Dettmers (Ai2) & Dan Fu (Together AI)
The MAD Podcast with Matt Turck
The podcast explores the debate around achieving Artificial General Intelligence (AGI), focusing on computational realities and practical applications. Tim Dettmers argues that advancements face diminishing returns due to physical constraints, suggesting hardware capabilities are nearly maxed out. Dan Fu counters that current hardware is underutilized, with potential for significant performance gains through optimization and new architectures. The discussion highlights the importance of post-training and usefulness, agreeing that focusing on practical applications and economic impact is more relevant than pursuing an abstract AGI. Both guests share insights on coding agents, emphasizing their potential to enhance productivity across various tasks, while also noting the challenges of integrating AI in education and the workforce.
Part 1: AGI Debate, Constraints, and Definitions
Part 2: Training Methods and Economic Impact
Part 3: Coding Agents and Practical Automation
Part 4: Technical Innovations and Future Outlook
Sign in to continue reading, translating and more.
Open full episode in Podwise
