The U.S. stock market and a resource-constrained parent gawp at the hottest new “thrifty” language model - but who can do more with less?

In this episode of The Jen AI Podcast, host Jenni Munroe explores the implications of a breakthrough in AI technology by Chinese company DeepSeek, which released an efficient language model that rivals Western large language models like ChatGPT despite resource constraints. Jenni draws humorous parallels between the AI news and the 1980s cult-classic Madonna movie 'Desperately Seeking Susan,' discusses key techniques that DeepSeek used (esp. model distillation and reinforcement learning), and assesses the broader impacts on the economy and individuals. She also reflects on her experiences using the DeepSeek model and other AI tools, while offering insights into optimizing resources in AI development and everyday life.

00:00 Introduction to the Jen AI Project

01:10 DeepSeek's Market Impact

03:31 "Desperately Seeking Susan" Movie Parallels

04:47 DeepSeek's Efficiency and Market Reactions

06:42 Model Distillation Controversy

10:30 Reinforcement Learning and AI Reasoning

15:00 The Legally Blonde Benchmark Test

18:09 Reflections and AI Tools Used

27:16 DOGE

--- Bonus material ---

28:08 NotebookLM's podcast on the DeepSeek research paper

29:52 Positive Reinforcement in AI Training

30:04 DeepSeeker's Self-Taught Reasoning

30:45 Majority Voting Boosts Accuracy

31:59 Polishing DeepSeek R1's Reasoning

33:41 Distillation: Making AI Accessible

37:11 Challenges and Future Research

44:05 Ethics and Impact of AI Reasoning

46:14 Conclusion: The Future of AI Reasoning

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit jenaihacks.substack.com