Jared Dunnmon, founder of a stealth startup and former DIU member focusing on AI, dives into the heated debate between open-source and scaled AI models. He warns of the potential risks posed by China's advancements in AI, emphasizing the need for the U.S. to maintain dominance in this critical technology. The discussion also covers the complexities of trusting open versus closed models, the importance of secure AI development, and how benchmarking and collaboration can help navigate these competitive waters.
39:02
forum Ask episode
web_stories AI Snips
view_agenda Chapters
menu_book Books
auto_awesome Transcript
info_circle Episode notes
question_answer ANECDOTE
Testing DeepSeek on Christmas
Jared Dunnmon started by interacting with DeepSeek's AI on Christmas Day to test its limits and censorship.
He discovered the AI refused to reveal its guardrails, prompting deeper investigation into Chinese AI censorship mechanisms.
insights INSIGHT
China's AI Market Share Strategy
China leverages low-cost chips and open AI models to gain market share rapidly.
This strategy risks funneling capital into Chinese tech ecosystems, challenging US competitors and AI leadership.
insights INSIGHT
Open Weight vs Open Source AI
Open weight AI models differ from open source by exposing only the final model, not the training data or code.
This limits transparency and complicates detecting hidden malicious behaviors within the model.
Get the Snipd Podcast app to discover more snips from this episode
Published in 1949, '1984' is a cautionary tale by George Orwell that explores the dangers of totalitarianism. The novel is set in a dystopian future where the world is divided into three super-states, with the protagonist Winston Smith living in Oceania, ruled by the mysterious and omnipotent leader Big Brother. Winston works at the Ministry of Truth, where he rewrites historical records to conform to the Party's ever-changing narrative. He begins an illicit love affair with Julia and starts to rebel against the Party, but they are eventually caught and subjected to brutal torture and indoctrination. The novel highlights themes of government surveillance, manipulation of language and history, and the suppression of individual freedom and independent thought.
Since the launch of Project Stargate by OpenAI and the debut of DeepSeek’s V3 model, there has been a raging debate in global AI circles: what’s the balance between openness and scale when it comes to the competition for the frontiers of AI performance? More compute has traditionally led to better models, but V3 showed that it was possible to rapidly improve a model with less compute. At risk in the debate is nothing less than American dominance in the AI race.
Jared Dunnmon is highly concerned about the trajectory. He recently wrote “The Real Threat of Chinese AI” for Foreign Affairs, and across multiple years at the Defense Department’s DIU office, he has focused on ensuring long-term American supremacy in the critical technologies underpinning AI. That’s led to a complex thicket of policy challenges, from how open is “open-source” and “open-weights” to the energy needs of data centers as well as the censorship latent in every Chinese AI model.
Joining host Danny Crichton and Riskgaming director of programming Laurence Pevsner, the trio talk about the scale of Stargate versus the efficiency of V3, the security models of open versus closed models and which to trust, how the world can better benchmark the performance of different models, and finally, what the U.S. must do to continue to compete in AI in the years ahead.