Episodes (Page 8)
✨
Ben Firshman (Replicate co-founder) traces company's 5-year journey from 2017 arXiv Vanity project to 2M+ users and $40M Series B
✨
Erik Bernhardsson (Modal founder, ex-Spotify) presents serverless infrastructure specifically designed for AI engineers' compute and latency requirements
✨
Ce Zhang (CTO) and Vipul Ved Prakash (CEO) of Together AI discuss the company's evolution and ~$100M Series A at $360-565M valuation
✨
David Hsu (Retool CEO) discusses the 50% decline in StackOverflow usage and its implications for developer productivity in the AI era
✨
Introduces 'The Four Wars of the AI Stack' framework: Data Quality Wars, GPU Rich vs Poor, Multimodality Wars, and RAG/Ops War
✨
Hugo Laurençon and Leo Tronchon (HuggingFace M4) explain building open source multimodal models by combining existing LLMs and vision encoders with adapter layers
✨
Nathan Lambert provides deep dive into RLHF (Reinforcement Learning from Human Feedback), explaining how transformer models transition from next-token prediction to helpful, honest assistants
✨
Steve Ruiz discusses tldraw, an AI canvas tool enabling sketch-to-working-website generation using GPT-4V for coding visual designs
✨
Mega episode featuring multiple startup founders recapping NeurIPS 2023 achievements, product launches, and industry trends
Top Startups
✨
Survey of 3,586 papers accepted to NeurIPS 2023, with audio guide commentary on standout papers and research directions
Best Papers
✨
Suhail Doshi builds Playground AI as AI-first image editor, advancing beyond rudimentary text-to-image tools from 2022-2023 era
✨
Beyang Liu and Steve Yegge from SourceGraph discuss 15+ years of code indexing and retrieval, teaching masterclass on code search history and interfaces
Beyang Liu
✨
Wing Lian's Axolotl is emergent finetuning library of choice for open models like Llama 2 and Mistral 7B, enabling customization and ownership
✨
Bryan Bischof frames notebooks as 'Chat++' interface
✨
Dylan Patel of SemiAnalysis breaks down GPU supply, production, and design landscape; coins 'GPU poor' vs 'GPU rich' divide affecting AI industry access
✨
Cleaned audio version of OpenAI DevDay coverage featuring spot interviews with Nvidia, Humanloop, Stealth, Zapier, MultiOn, Bloop.ai, and Guardrails representatives
✨
OpenAI DevDay delivered GPT-4 Turbo with expanded context, custom GPTs with knowledge/actions, and a no-code builder
✨
Phind combines code search with LLM assistance to help developers find and implement technical solutions more effectively than traditional search
✨
Artem Keydunov of Cube.dev discusses text-to-SQL as a commodity application now with multiple ChatGPT plugins and competing startups entering the space
✨
Jeremy Howard from Fast.ai discusses how finetuning paradigms are shifting with larger context windows and better base models reducing the need for traditional finetuning