The Download: reawakening frozen brains, and the AI Hype Index returns
A decade-frozen brain could aid organ transplants, while Niantic's Pokémon Go data builds a real-world AI map.
This edition of The Download from MIT Technology Review highlights the intersection of long-term scientific ambition and practical AI application. In a notable cryonics story, the brain of L. Stephen Coles has been preserved at -146°C for over a decade. His friend, cryobiologist Greg Fahy, is researching its potential for future reanimation, a goal met with skepticism. However, the work could yield tangible benefits by advancing techniques for cryopreserving organs for transplantation, moving the technology closer to medical reality.
On the AI front, Niantic Spatial—a company spun out from Pokémon Go creator Niantic—is leveraging a unique asset: the massive, crowdsourced AR data from over 500 million app installs. The company is using this data to build a precise 'world model,' a foundational technology that grounds large language models (LLMs) in real-world environments. The immediate application is to provide delivery robots with inch-perfect navigation capabilities, turning a gaming phenomenon into a tool for robotics.
The newsletter also marks the return of the AI Hype Index, a tool designed to separate industry reality from fiction. Additional coverage includes a subscriber discussion on the next era of space exploration and news briefs on OpenAI shutting down its Sora video app, a legal battle between Anthropic and the Pentagon, and Arm's move to sell its own AI chips.
- Cryobiologist studies friend's brain frozen for 10+ years at -146°C, with potential applications for organ transplant preservation.
- Niantic Spatial uses Pokémon Go's AR data from 500M+ installs to build a 'world model' for precise robot navigation.
- The AI Hype Index returns to help professionals track real trends amid industry noise, alongside major news from OpenAI and Arm.
Why It Matters
Professionals can track real AI progress beyond hype while seeing how unconventional data (like game maps) solves hard robotics problems.