Research papers, news, and signals from the field — curated and summarized.
Researchers are adapting large language models for languages with limited training data, opening up AI tools to billions who previously lacked access.
New satellite-based ML systems can now predict wildfire spread with 85% accuracy 6 hours in advance, enabling proactive evacuation coordination.
A landmark initiative will bring AI-powered diagnostic imaging and disease screening to clinics with limited physician access across Sub-Saharan Africa and Southeast Asia.
A randomized controlled trial of 3,000 participants found that AI-assisted CBT chatbots reduced mild anxiety symptoms at rates comparable to in-person therapy.
A growing community effort to catalog real-world AI deployments with social impact, track outcomes, and surface the most effective interventions.
A new vision model trained on labeled environmental data can now detect illegal waste sites, mine tailings, and pollution sources at global scale.
An 18-month pilot using personalized AI tutoring in low-connectivity rural schools shows significant gains in numeracy and literacy among grades 3–6.
Fintech lenders using mobile behavior, utility payments, and social data are reaching borrowers completely invisible to traditional credit bureaus.
A multimodal model trained on signed language corpora can now interpret and generate sign language in real time, removing communication barriers in education and employment.
Three governments have quietly piloted AI tools to draft clearer, more consistent legislation — raising important questions about accountability and public oversight.