Ep 47: Chief AI Scientist of Databricks Jonathan Frankle on Why New Model Architectures are Unlikely, When to Pre-Train or Fine Tune, and Hopes for Future AI Policy
In this podcast, Jonathan Frankle, Chief AI Scientist at Databricks, dives into enterprise AI strategies, covering topics like model selection—whether to use prompt engineering, fine-tuning, or training—and the significance of thorough evaluations. He advocates for an iterative approach, suggesting that businesses start small with accessible models and scale up as they see clear returns on investment. Frankle also discusses the current landscape of AI research, stressing the value of human-in-the-loop systems and the necessity for responsible AI development, especially in critical fields like healthcare and autonomous vehicles. He underscores the vital contribution of academia in advancing the field.
