The Data Scientist Show - Daliana Liu Case studies from the GenAI frontier, scaling ML teams, from biologist to machine learning consultant- Erik Gafni - The Data Scientist Show #082
Feb 24, 2024
Discussion on GenAI projects, stable diffusion models for social media apps, AI in biotech, and scaling ML teams. Insights on self-supervised learning, research vs production, AGI, and data quality in GenAI. Erik's journey from biologist to ML consultant, mistakes made, and new trends in GenAI. Philosophy in LLMs, OpenAI vs Open Source, and how he hires people.
AI Snips
Chapters
Transcript
Episode notes
GenAI's Reality Check
- Generative AI is accelerating rapidly culturally and technologically, but some expectations are premature.
- Current large language models can't yet fully automate complex cognitive jobs, but consumer chatbots show early promise.
Building Real-Time Accent AI
- At SONUS, they built real-time accent removal by refactoring code and introducing MLOps practices.
- They streamlined workflows with DAXTER, improving scientists' productivity by automating tasks.
Prioritize Data Quality
- Spend significant time on data quality assessment; it can reveal hidden issues.
- Manually inspecting and cross-validating with others improves overall data quality.