

DeepSeek Disrupts Everything
Jan 31, 2025
In this discussion, Keith Brown, a talented voice artist and audiobook narrator, shares insights from his work on Perry Carpenter’s book, FAIK. The conversation dives into the rise of DeepSeek, a groundbreaking Chinese AI that's stirring up the tech world. They explore its innovative training methods, geopolitical implications, and the challenges it faces against Western rivals. Other intriguing topics include the double-edged sword of AI in artistic expression and the evolving challenges of social engineering in a digital age.
AI Snips
Chapters
Books
Transcript
Episode notes
DeepSeek's Disruptive Impact
- DeepSeek, a Chinese open-source AI model, disrupts the AI landscape by achieving comparable performance to GPT-3.5 at a significantly lower cost.
- This challenges the notion that bigger budgets always yield better results, highlighting the importance of optimization.
DeepSeek's Training Cost
- DeepSeek's training reportedly cost $5.5 million, significantly less than other large language models (LLMs).
- This figure may only represent the final training run and not the full R&D cost, possibly leveraging existing models like GPT-4.
DeepSeek's Training Method
- DeepSeek utilizes reinforcement learning without supervised fine-tuning, comparing its output to a "teacher" model.
- Evidence suggests DeepSeek may have used ChatGPT as its teacher model, potentially violating terms of service.