The Right Way to Worry About AI
Scientists at Stanford University and the Arc Institute trained an AI model named Evo to recognize patterns in naturally occurring DNA. The model was then able to extrapolate these patterns to create functional, never-before-seen viruses. Researchers used these DNA sequences to genetically modify bacteria, which produced the novel viruses that were subsequently able to infect other bacteria, proving their viability.
This research demonstrates the potential for AI to design novel biological threats, raising significant biosecurity concerns and highlighting the urgent need for proactive regulation and robust technical guardrails to manage dual-use AI dangers.
The episode also covers OpenAI's new model Astro and pricing changes, Stripe's acquisition talks for OpenRouter, NVIDIA's potential memory shortages for its Rubin GPUs, OpenAI's rumored smart speaker device, and SoftBank's $10 billion loan secured against its OpenAI stake.
AI BIOSECURITY EVO AI MODEL NOVEL VIRUSES BIODEFENSE AI REGULATION