The world, the universe and us Weekly: Is DeepSeek really the ChatGPT killer?; alarming scale of ocean warming; dolphin peeing contests
15 snips
Jan 31, 2025 In this discussion, Jeremy Hsu, a reporter for New Scientist, dives into the rise of DeepSeek, a Chinese AI model that rivals ChatGPT at a fraction of the cost. He explains the innovative technologies behind DeepSeek and its implications for the AI arms race. The conversation also tackles the alarming rise in ocean temperatures, attributed to accelerated heat absorption. Lastly, there's a light-hearted exploration of Amazon river dolphins engaging in an unusual urine projection behavior, shedding light on their unique social interactions.
AI Snips
Chapters
Transcript
Episode notes
DeepSeek's Efficient AI
- DeepSeek's new AI model, R1, rivals leading U.S. models in performance, particularly in math and coding.
- Surprisingly, it was developed at a fraction of the cost, challenging the "scale is all you need" mantra.
DeepSeek's Strategic Release
- DeepSeek's release of R1 might be a strategic move by the Chinese government to showcase their AI progress.
- This coincides with Trump's promotion of a $500 billion AI initiative, suggesting a potential "Sputnik moment."
OpenAI's Allegations
- OpenAI alleges DeepSeek trained R1 using ChatGPT outputs, evidenced by R1 sometimes responding like ChatGPT.
- Ironically, OpenAI faces similar accusations of using writers' work without permission to train ChatGPT.