$1.8 billion bet on the data behind AI models of biology
Biohub, the US Department of Energy, the NIH and new partners are committing $1.8 billion in funding, data, compute and measurement technology to build an open, AI-ready biological dataset, which they call the largest coordinated effort of its kind. DOE puts in more than $500 million over five years through its Genesis Mission, using exascale supercomputers and national lab facilities. NIH will standardise datasets from more than $500 million of earlier federal work for model training. Google DeepMind, Isomorphic Labs and Meta are adding $300 million to the Virtual Biology Initiative. The bottleneck for a "virtual cell" has been data, not model architecture, and this is the first serious attempt to pay for that at national scale. Biohub via PR Newswire
What else matters
ChatGPT gets "Intelligent UI" with GPT-6
GPT-6 is rolling out across ChatGPT with Intelligent UI, which answers with interactive charts, calculators, editable graphs and tappable buttons instead of plain text. Plus, Pro, Business and Enterprise got it Wednesday; Free and Go tiers follow Thursday. Users can dial the visuals back if they prefer text. TechCrunch
Google launches Playground for prompt-built games
Google's new experimental Playground lets people create, play and share games from text prompts with no coding. It's live for US users aged 18 and over, with creation tiers tied to Google One membership. A Unity Spark integration for higher-fidelity 3D games is in testing. Unite.AI
Follow-up: Mistral Large 4 preview
After yesterday's reveal, Mistral's trillion-parameter Large 4 is in public preview, with Mistral claiming it matches or beats leading Chinese open-weight models in several areas. Security specialists and state authorities get early access; open weights are still due at the end of October. Mistral
