The a16z Show When Will AI Hit the Enterprise? Ben Horowitz and Ali Ghodsi Discuss
117 snips
Oct 6, 2023 
Join Ben Horowitz, co-founder of a16z, and Ali Ghodsi, CEO of Databricks and a leader in open-source data technologies, as they dive into the challenges of integrating AI in enterprises. They discuss the internal politics and data ownership issues large organizations face. The duo also explores the tug-of-war between large language models and specialized models, emphasizing efficiency and accuracy. Ethical implications weave throughout the conversation, reflecting on historical missteps and the need for balance in future AI advancements.
AI Snips
Chapters
Transcript
Episode notes
Building Proprietary Models
- Enterprises with specific use cases can build better models with their own data using tools like Mosaic.
- This allows for complete data ownership and control.
Model Size and Efficiency
- For specific enterprise use cases, smaller, fine-tuned models are often more efficient.
- Larger models are costly to run, even if training is affordable.
The Future of Large Models
- Databricks is focused on building large models to explore fine-tuning techniques for specific tasks.
- The ideal future involves a large base model with specialized, efficient add-ons.