Anthropic CEO Dario Amodei’s signature prediction — that AI would compress 50-100 years of biological progress into 5-10 — now carries a revised timeline: maybe a decade from now, meaning 2036. At Anthropic’s “The Briefing: AI for Science” event in San Francisco on June 30, Amodei didn’t declare victory; he bought time, pushing the “compressed 21st century” out of the near-term hype window.
Claude Science: Chatbot Meets Lab Bench
Claude Science, launching in beta, is the tangible outcome of that vision. Alexander Tarashansky, who led development, demoed a product that feels like a chatbot but packs richer tools: real-time infographic generation, access to more than 60 scientific databases, and smart correction of visualization flaws via natural language annotations. Eric Kauderer-Abrams, Anthropic’s head of life sciences, put it simply — “Science is a very visual affair.” The demo showed Claude Science fixing hard-to-read labels and repositioning legends after brief user comments. That capability, if it reaches mainstream Claude, would drag AI assistants beyond endless text exchanges.
The ‘Cure Cancer’ Trap and Realistic Timelines
Every AI science discussion inevitably hits “cure cancer.” Bristol Myers Squibb CEO Chris Boerner pushed back: “We don’t want to get over our skis.” He acknowledged substantial progress but warned against promising delivery within a single lifetime. GLP-1 pioneer Lotte Knudsen pointed out a hard constraint: even perfect AI can’t shrink a five-year clinical trial to zero. “It’s probably hard to imagine that you can go below five years,” she said. Novartis CEO Vas Narasimhan offered a concrete number: AI might cut drug development from 12 years down to 7-8 years. That’s not a 10X speedup, but compounded across the industry, it’s massive.
Hallucinations as Feature, Not Bug
Interviewer Matthew Herper admitted he’d asked Claude for debate questions, and the bot suggested, “Why should pharma trust AI predictions when your models hallucinate?” Amodei’s response was refreshingly honest: hallucinations have improved but will never fully disappear, because creativity itself straddles the boundary between making things up and having good ideas. He compared it to human cognition — we don’t expect people to stop brainstorming. He didn’t directly address what that means for scientific rigor, but the willingness to entertain the criticism on his own stage was notable.
The endgame isn’t a single miracle drug. Knudsen’s call for “bilingual” scientists — fluent in both a scientific domain and AI — points to the real bottleneck. One person on each team who can translate between wet lab and neural net might do more than any model release.
Source: 5 takeaways from Anthropic's big science event
Domain: fastcompany.com
Comments load interactively on the live page.