How a $6 Million Model Upended a Trillion-Dollar Industry | Ramesh Raskar | TEDxBoston

What happens when a little-known AI lab in China builds a groundbreaking model for less than the cost of a Hollywood movie—and then gives it away for free? In this eye-opening talk, MIT Professor Ramesh Raskar—pioneer of AutoML (AI that writes AI), privacy-preserving machine learning, and decentralized AI—unravels how DeepSeek’s $6 million breakthrough triggered a $1 trillion tech sell-off, shattered assumptions about AI’s “moonshot economics,” and exposed Silicon Valley’s most overlooked vulnerability: decentralization. From the geopolitics of open-source AI to the paradox of doing more with older chips, Raskar—whose innovations at MIT enable AI to work across data silos, untrusted environments, and agentic systems—explores: 🔹 Why constraints like embargoed tech spark radical innovation 🔹 How decentralized AI could upend power structures, inspired by his research on data markets and agentic systems 🔹 The rise of “small AI” and what it means for startups, nations, and the future of privacy 🔹 Silicon Valley’s reckoning: When efficiency beats brute-force spending This isn’t just a story about algorithms—it’s about power. For policymakers, investors, and anyone curious about AI’s future, Raskar reveals why the next revolution in artificial intelligence might not come from who (or where) you expect… and why decentralized systems could rewrite the rules of global tech. Ramesh Raskar is an MIT professor and founder of MIT’s Decentralized AI initiative. His pioneering work in AI includes: AutoML (AI that designs better AI, 2015), Privacy-preserving ML (AI across data silos, 2017), Data Markets (pricing data for AI, 2021), Decentralized agentic systems (AI collaboration in untrusted environments, 2022). He delivered a popular TED talk on trillion-frame-per-second imaging. Winner of the National Academy of Inventors Award (2024) and the Lemelson Prize (2016, considered the “Nobel of Engineering”), he has led AI teams at Google X, Apple, and Meta during MIT sabbaticals. He has founded multiple startups and nonprofits and co-founded C10 Labs, a AI Venture Studio. He holds 100+ US patents in machine learning, imaging systems, and health tech. He teaches the most popular AI Venture class jointly taught at MIT and Harvard. This talk was given at a TEDx event using the TED conference format but independently organized by a local community. Learn more at https://www.ted.com/tedx