Explore how recent advancements in large language models are challenging established machine learning principles in this 27-minute talk by Emmanuel Ameisen from MLOps.community. Examine which traditional ML rules may become obsolete and which remain crucial as models develop capabilities like self-supervision, extrapolation, and unprecedented scale. Learn about emerging practices such as models generating their own training data, self-evaluation, and achieving strong performance in new domains. Gain insights into the changing landscape of ML methodology and its implications for future model design and evaluation approaches.
Challenging Traditional Machine Learning Principles with LLMs - Part 2
Overview
Syllabus
Intro
ML Fundamental Fundamentals
Simple Models
Use Relevant Data
Use the Internet
Relevant Data
Model Generated Data
Training Data
Constitutional AI
Improvement
Evaluation
Purpose
Outro
Taught by
MLOps.community
