This Week in Startups

Expanding AI chip capabilities beyond Nvidia with Modular CEO Chris Lattner | E1808

10 snips
Sep 14, 2023
Guest
Chris Lattner
Chris Lattner, CEO of Modular and a veteran of Apple, Tesla, and Google, dives deep into the evolving landscape of AI hardware. He discusses the challenges in deploying AI models and the complexities of the current technology stack. Lattner explains how Nvidia's dominance shapes the market and what Modular is doing differently. He also emphasizes the importance of collaboration in AI development and predicts a future filled with innovative, specialized applications that could disrupt the status quo.
Ask episode
AI Snips
Chapters
Transcript
Episode notes
INSIGHT

AI Hardware Utilization

  • AI hardware utilization is low during training due to time-sharing systems.
  • These systems, similar to HPC, allocate resources to researchers for varying durations, leading to idle periods.
INSIGHT

Training vs. Production

  • AI training and production are distinct, requiring different approaches.
  • Training uses supercomputers for large batch jobs, while production needs scale-out systems for many users.
ANECDOTE

ML Deployment Challenges

  • Deploying ML involves using various point solutions like Apple’s CoreML, creating friction.
  • Hardware makers prioritize their own solutions, neglecting cross-platform compatibility.
Get the Snipd Podcast app to discover more snips from this episode
Get the app