Astera Institute Commits $500M to AI-Ready Biology as AI Tools Disrupt Science Publishing
The Astera Institute invests $500M in Radial to build AI-friendly biology infrastructure. Meanwhile, AI tools flood computer science publishing, AI-assisted NIH grants lean toward less innovation, and a new platform lets AI agents hire human helpers.
The Astera Institute has committed $500 million to Radial, its Emeryville, CA-based life sciences division, with a mission to reimagine data generation, sharing, infrastructure and tools necessary to create the best conditions for AI to help solve biological problems. Becky Pferdehirt, CEO of Radial and former investor at Andreesen Horowitz’s Bio & Health Fund, described the work as "unglamorous, unsexy infrastructure and tools."
AI tools are reshaping scientific research, sometimes with disruptive effects. In computer science, tools that write code, draft papers and automate research steps have turbocharged the production of preprints and conference submissions, many of which are fake or of low-quality. This flood has overwhelmed reviewers, raising concerns about trust in the field. "Trust in scientific research, particularly within computer science, faces a substantial risk of erosion," said computer scientist Seulki Lee.
AI's influence extends to research funding. Grant proposals submitted to the US National Institutes of Health that were drafted with AI assistance are consistently less distinct from previous research than proposals written without AI, and are slightly more likely to be funded. Preliminary data suggest that AI might pull research toward safe, less-innovative ideas. However, science-policy researcher Cassidy Sugimoto noted that the findings don't necessarily mean AI is fundamentally shifting which ideas are funded; researchers might be using AI to better communicate their own ideas.
In a different melding of AI and human work, a platform called RentAHuman.ai now allows AI agents to hire people for tasks they cannot accomplish alone. Human users—referred to as "meatspace workers"—create profiles advertising skills (such as attending meetings or taking photos) and set their expected pay. Software engineers Alexander Liteplo and Patricia Tani launched the platform after Liteplo 'vibe coded' the system in about a day and a half.