This Day in AI Podcast Llama 2 Unleashed, Open-Source Vs OpenAI, Instructions for ChatGPT & Fine Tuning with Llama2 | EP24
19 snips
Jul 21, 2023 The podcast discusses the release of Llama 2, a commercially available open-source AI model. They explore how open-source models threaten the business models of OpenAI and others. Other topics include Instructions for ChatGPT, fine-tuning smaller models, Google's Generative AI Enterprise Search product, Bing Enterprise Chat, AppleGPT rumors, and the state of AI startups and VC funding.
AI Snips
Chapters
Transcript
Episode notes
Fine-Tune Small Models On Quality Data
- Fine-tune smaller models on high-quality domain data to solve specific tasks cheaply.
- Host the tuned model yourself to avoid variable API costs and retain privacy.
Hosts Found Llama 2 Surprisingly Good
- Chris Sharkey reported being impressed by Llama 2's quality and felt it sometimes beat GPT-3.5.
- He tested Llama 2 personally and found its responses strong in his interactions.
Mix Models To Fit Context And Cost
- Multi-model strategies let you mix small, specialized models with larger ones for cost and performance.
- Use embeddings and retrieval to handle large context with smaller context-window models.