Tag
A new research paper introduces ASI-Arch, an autonomous AI system capable of discovering novel neural network architectures without human-designed search spaces. By running thousands of automated experiments, it generated over 100 new state-of-the-art linear attention models, signaling a major shift toward AI-driven scientific collaboration.
This paper presents an agentic system using Large Language Models to automate the discovery of exchange-correlation functionals in Density Functional Theory, achieving improvements over human-designed baselines while highlighting challenges with benchmark overfitting.
This paper introduces AeroJEPA, a Joint-Embedding Predictive Architecture for scalable 3D aerodynamic field modeling. It addresses limitations in current surrogate models by predicting semantic latent representations of flow fields, enabling efficient high-fidelity analysis and design optimization.
OpenProtein.AI, founded by MIT researchers Tristan Bepler and Tim Lu, has launched a no-code platform to democratize access to advanced AI models for protein design and engineering among biologists.
MIT researchers published a paper in 'Matter' describing an AI model that uses noninvasive neutron-scattering data to classify and quantify atomic defects in materials. The model can detect multiple defect types simultaneously, improving the characterization of semiconductors and other materials without damaging them.
Google DeepMind announced a new partnership with the Indian government to accelerate scientific discovery and education through AI, including providing access to models like AlphaGenome and launching a $30 million Impact Challenge.
OpenAI and the U.S. Department of Energy have signed a memorandum of understanding to collaborate on AI and advanced computing initiatives, including the Genesis Mission, aiming to accelerate scientific discovery through frontier AI models deployed in real research environments.
Google DeepMind announces a strengthened partnership with the UK government to deploy frontier AI models like AlphaEvolve and AlphaGenome for scientific discovery, education, and national security. The collaboration also includes plans to establish DeepMind's first automated science laboratory in the UK focused on materials science.
Google DeepMind is partnering with the U.S. Department of Energy to support the Genesis Mission, providing scientists access to AI tools like the AI co-scientist to accelerate scientific discovery and innovation.
Google DeepMind and Yale released C2S-Scale, a 27B parameter foundation model built on Gemma for single-cell analysis that discovered a promising drug combination (silmitasertib and interferon) to enhance immune visibility of "cold" tumors, with predictions validated through experimental confirmation.