Thinking Machines: AI & Philosophy The Future is Fine Tuned (with Dev Rishi, Predibase)
7 snips
May 24, 2024 Dev Rishi, CEO of Predibase, discusses the future of AI fine-tuning, challenges in model hosting, and competition with OpenAI. They explore the benefits of multi-modal language models, optimization methods like Lorax, and the advancements in audio modeling. The conversation also touches on balancing work and personal life in the AI industry.
AI Snips
Chapters
Transcript
Episode notes
Commoditization of Model Hosting
- Model hosting is becoming commoditized, making cost a primary differentiator.
- Focus on performance and advanced workflows to avoid this commoditization.
LoRA: Efficient Fine-Tuning
- LoRA, a parameter-efficient fine-tuning method, customizes a small subset of model weights.
- This allows for task-specific training without overwriting foundational knowledge.
LoRAX Origin Story
- Predibase developed LoRAX to enable a free trial for their fine-tuning product.
- The cost of training was negligible, but hosting individual fine-tuned models was a concern.