Skip to main content
AmigoFish

How China Changed Its Views on AI Safety and AI's Impact - with Matt Sheehan

From The AI Power Podcast · · 1 hr 1 min

This plays from the publisher's own server. We do not host a copy.

Open at the source

Not rated yet

Description

Matt Sheehan joins Gregory Allen to talk about how the Chinese government views AI safety and how those views have evolved over the past year. They discuss notable policy documents that give an insight into how key AI decision makers view the risks that arise from AI’s developments, the effect that external events like Mythos or the HuggingFace incidents had on AI safety views, and lessons to draw from past U.S.-China AI dialogues. Then, Greg and Matt discuss the promise of a “parallel” AI safety engagement, alongside the effect of Chinese export controls on diffusion and open models.

CHAPTERS: 00:00 - Welcoming back Matt Sheehan, the China AI policy research community's "Elder Statesman" 1:41 - 9:43: What institutions matter in Chinese AI Policy discussions - CAC and NDRC 9:43 - 15:32: Who should the U.S. talk to in the Chinese government when it comes to AI safety? 15:32 - 30:19: How is China thinking about AI safety? How has that changed? 30:19 - 35:18: What effect did Mythos’ release and the HuggingFace incident have on China’s view of AI safety? 35:18 - 39:40: AI safety and U.S.-China diplomacy failures: the 2024 talks 39:40 - 47:27: Bona fide AI safety cooperation or pursuing AI AI safety in parallel? Plus, previewing upcoming U.S.-China AI dialogues 47:27 - 52:48: China’s views on open source models and AI export controls 52:48 - 1:01:06: The Chinese government’s plans to deal with AI’s impact on job security and employment

Matt’s work at the Carnegie Endowment: https://carnegieendowment.org/people/matt-sheehan Twitter: https://x.com/mattsheehan88 Read more on Matt Sheehan’s Substack: https://mattsheehan.substack.com/

Follow Greg on Twitter https://x.com/Gregory_C_Allen?s=20 Learn about Decision Tree Research at https://decisiontreeresearch.com

Back to The AI Power Podcast · Report a problem