The MAD Podcast with Matt Turck Can AI Infrastructure Work Like Magic? | Erik Bernhardsson, CEO, Modal
19 snips
Oct 31, 2024 
Erik Bernhardsson, CEO of Modal and former Spotify developer, shares insights on transforming the AI infrastructure landscape. He discusses the shift from training to inference as a primary focus and highlights Modal's goal of simplifying the developer experience. Erik explains the challenges of GPU accessibility and the emergence of specialized AI cloud providers. He also envisions a future filled with AI-native consumer applications, emphasizing the importance of creating products that developers love and efficient feedback loops.
AI Snips
Chapters
Transcript
Episode notes
Suno Use Case
- Suno uses Modal for generating AI music, training their own models externally.
- This exemplifies Modal's focus on handling infrastructure while users focus on model building.
Serverless for Data/AI/ML
- Serverless computing did not catch on with backend web development, but it suits data/AI/ML applications.
- These applications often have bursty workloads and don't always require hyper-low latency.
Cloud Maximalism
- Modal has been cloud maximalist, prioritizing multi-tenancy for better resource pooling and cost optimization.
- They believe a superior product can overcome customer concerns about multi-tenancy.