Source linked

Claude Science trouve de manière autonome des candidats médicamenteux pour la phénylketonurie

technologyreview.com@market_structureyesterday·Artificial Intelligence·6 comments

Le nouveau produit phare d'Anthropic identifie de manière autonome les candidats aux médicaments pour les maladies génétiques rares, positionnant l'entreprise comme un défi sérieux à la domination décennale de DeepMind dans l'IA pour la science.

anthropicclaude sciencedeepminddrug discoveryai for sciencejohn jumper

Claude Science—Anthropic's new flagship product—autonomously identified new drug candidates for phenylketonuria during a live demo for pharma executives on Tuesday.

How Claude Science Differs from Claude Code

Claude Science is a full-featured standalone product, elevated to the same rank as Claude Code and Claude Cowork. Eric Kauderer-Abrams, Anthropic's head of life sciences, said it “represents how important this is to our mission.” Unlike the earlier “Claude for Life Sciences” plug-ins, Claude Science writes code, runs it on powerful computer clusters, and prioritizes reproducibility so every figure can be traced back to its source. It can interface with tools in genetics, chemistry, and protein biology—all designed to accelerate drug development.

Why Anthropic Is Betting on Drug Discovery

Pharmaceutical companies have deep pockets, and Anthropic is set to see its first profitable quarter this year with an IPO approaching. The company isn’t just licensing the tool: it's using Claude Science to pursue its own research into neglected diseases, both to advance science and to stress-test the product in the real world. Alexander Tarashansky, who led development, demonstrated the system autonomously hunting for phenylketonuria drug candidates—a rare genetic disease that would be uneconomical for most pharma R&D pipelines.

The DeepMind Challenger

For a decade, DeepMind owned AI for science—AlphaFold won John Jumper and Demis Hassabis a Nobel Prize. But Jumper left DeepMind for Anthropic earlier this month. Meanwhile, Harvard physicist Matthew Schwartz estimated that Anthropic’s Opus 4.5 model is about as capable as a second-year graduate student on scientific projects. With CEO Dario Amodei holding a PhD (unlike OpenAI’s Sam Altman), Anthropic is well positioned to take up DeepMind's mantle—and Claude Science is the product that makes that play explicit.

If Claude Science lands big pharma contracts, it could help Anthropic stay profitable as its IPO approaches later this year.

Source: Claude Science is Anthropic's newest flagship product
Domain: technologyreview.com

Read original source ->

External source stays available while the OJO article and comment thread stay local.

More in Artificial Intelligence

view topic
Fable 5 Back Online, Still Happy to Plan Botnets For You

Anthropic's re-released Fable 5 model immediately fell for a basic dual-use prompt trick, planning IoT botnet attacks - while GPT-5.5 and Opus 4.8 refused.

Gemma 4 12B Runs Locally on 16GB Laptops: Google's June AI Blitz

Gemma 4 12B brings vision, voice, and reasoning to laptops with 16GB memory, while Gemini 3.5 Live Translate handles 70 languages in real time.

HippoRAG Runs Personalized PageRank on Neptune for Single-Pass Multi-Hop QA

AWS implementation hits 0.8234 PageRank score on a Leonardo da Vinci multi-hop query, proving single-pass graph retrieval can replace iterative processes.

Gemma 4 on Cerebras Cuts Voice AI Latency Below Human Perception Threshold

Hugging Face and Cerebras built a speech-to-speech pipeline that achieves stable sub-second response times using Gemma 4 31B, Nvidia Parakeet, and Qwen3TTS, already powering 9,000 robots.

Comments load interactively on the live page.