Kyle Corbett and Daniel from CoreWeave (formerly Openpipe) discuss the practical advantages of Reinforcement Learning (RL) over Supervised Fine-Tuning (SFT) for building reliable and efficient AI agents. They introduce Serverless RL, a new platform designed to eliminate the infrastructure complexities of RL training, and share a playbook for teams looking to get started.
A discussion on OpenAI's new browser ChatGPT Atlas, Andrej Karpathy's pessimistic timeline for AI agents, the DeepSeek-OCR paper on visual context compression, and a study suggesting large language models can suffer from "brain rot" when trained on low-quality social media data.
Amjad Masad, founder of Replit, joins a16z to discuss the rise of AI agents that can now plan, reason, and code for hours. He explains how reinforcement learning and verification loops unlocked long-horizon reasoning, why AI is advancing fastest in verifiable domains like code, and debates whether "good enough" AI might be a local maximum that blocks the path to AGI.
A breakdown of Machine Learning (ML), its relationship with AI and Deep Learning, and its core paradigms: supervised, unsupervised, and reinforcement learning. The summary explores classic models and connects them to modern applications like Large Language Models (LLMs) and Reinforcement Learning with Human Feedback (RLHF).
Jason Droege, CEO of Scale AI, discusses the evolution of AI training from simple labeling to complex, expert-driven tasks. He shares insights on the future of AI agents, the reality of enterprise AI adoption, and crucial business lessons learned from building Uber Eats from zero to a multi-billion dollar business.
A reflection on Richard Sutton's "Bitter Lesson," arguing that while his critique of LLMs' inefficiency and lack of continual learning is valid, imitation learning is a complementary and necessary precursor to true reinforcement learning, much like fossil fuels were to renewable energy.