Anthropic just unveiled new research on tracing the thoughts of large language models, building a "microscope" to inspect what actually happens inside AI networks. The tool helps researchers unpack Claude's complex and surprising inner mechanisms, making black-box models far more interpretable.
Dartmouth researchers just published the first clinical trial of a generative AI therapy chatbot, Therabot, showing participants experienced a 51% drop in depression symptoms and a 31% drop in anxiety. Patients trusted the bot almost like a human therapist, often reaching out during late-night crises. Why it matters: With 1,600 patients per provider in the U.S., clinically backed AI could bridge massive shortages, though researchers stress expert supervision remains essential.
ARC Prize just unveiled ARC-AGI-2 and launched its 2025 competition to stress-test general fluid intelligence. Every test task was solved by multiple humans in two tries or less, yet pure LLMs score a clean 0% and public reasoning systems barely reach single digits. Why it matters: simply scaling up parameters won't close this reasoning gap; AGI needs fundamentally new architectures.
Midjourney just teamed up with NYU researchers to tackle AI's biggest creative flaw: bland, repetitive writing. By introducing two new training methods called DDPO and DORPO, they reward rare and unique high-quality responses during fine-tuning on models like Llama 3.1. It turns out the image generator has serious ambitions for text, too.
Sakana AI just announced that a scientific paper fully generated by its upgraded The AI Scientist-v2 passed peer review at an ICLR workshop. Conducted with UBC and Oxford, it marks the first time a completely AI-authored paper cleared the exact same review process human researchers face. Full details of the v2 system will drop soon.
Luma just introduced Inductive Moment Matching (IMM), an open-source pre-training technique designed to replace conventional diffusion models. By taking an inference-first approach and incorporating target timesteps with moment matching, IMM delivers better image quality than diffusion or flow matching while cutting sampling steps by up to 30x.
Researchers just introduced Generalized Interpolating Discrete Diffusion (GIDD) to overcome the rigid limitations of next-token prediction. By blending masking with uniform noise, it achieves top-tier generation performance while finally allowing models to revise and self-correct their own mistakes.
Google just expanded its medical AI system AMIE from one-off diagnoses to ongoing disease management across multi-visit consultations. Powered by a dual-agent Gemini setup grounded in clinical guidelines, AMIE matched or outperformed primary care doctors across 100 simulated patient cases. It marks a crucial step toward AI-assisted chronic care, though real-world clinical validation is still pending.
OpenAI just launched NextGenAI, pledging $50M in research grants, compute funding, and API access across 15 leading institutions including MIT, Harvard, and Oxford. The initiative aims to fast-track discoveries in healthcare, metascience, and rare manuscript digitization. It is OpenAI's biggest academic push yet to embed its tech inside top labs.
UC San Diego’s Hao AI Lab just pitted top AI models against Super Mario Bros. using a custom framework. Anthropic’s Claude 3.7 took the top spot, while reasoning heavyweights like OpenAI’s o1 flunked out because taking several seconds to "think" in a platformer means instant death.
Meta researchers just unveiled Avat3r, an AI system that generates an animatable, photorealistic 3D avatar head in minutes using only four phone selfies. It uses a vision transformer to predict 3D Gaussians, drastically lowering the hardware barrier from specialized multi-camera rigs. While rendering currently crawls at 8 FPS on an RTX 3090, it points toward effortless avatar setup for future headsets.
OpenAI just teamed up with nine U.S. national labs to host a massive AI Jam Session, putting its frontier models into the hands of over 1,000 researchers across institutions like Oak Ridge and Berkeley. Scientists are stress-testing the models on complex domain-specific problems, with findings set to shape future research-focused AI systems in an upcoming report.
Meta just unveiled Aria Gen 2, its next-generation smart glasses built for machine perception and robotics research. The 75-gram frames run SLAM and multi-modal tracking entirely on-device via custom silicon, packing new nosepad sensors to isolate the wearer's voice and pulse. Why it matters: It provides AI researchers with a wearable testbed to collect egocentric data for training embodied AI and assistive tech.
Snap Research just introduced Set and Sequence, a framework for personalizing text-to-video models using just one video clip. By splitting dynamic concepts into an appearance frame set and a motion sequence via two-step LoRA encoding, it lets DiT models capture both how a subject looks and moves, making character cloning far more flexible.
Microsoft just open-sourced BioEmu-1, a deep-learning model designed to capture the structural changes proteins undergo in nature. Generating thousands of structures per hour on a single GPU, it slashes computational needs by up to 100,000x compared to standard molecular dynamics simulations. Why it matters: Modeling dynamic structural ensembles instead of static snapshots is crucial for designing better therapeutics.
Google's AI tool "co-scientist" just solved a complex superbug mystery in two days that took Imperial College London researchers an entire decade to prove. Given a short prompt, the AI independently proposed the team's unpublished hypothesis on antibiotic resistance, alongside four other viable theories the scientists are already investigating.
Arc Institute, Stanford, and NVIDIA just rolled out Evo 2, the largest publicly available biological foundation model. Trained on nearly 9 trillion nucleotides across all domains of life, it handles 1-million-token contexts and scored 90% accuracy in predicting breast cancer gene mutations.
Microsoft just unveiled Majorana 1, the world’s first quantum processor built on a topological core designed to scale up to one million qubits on a single chip. By using a newly engineered material to control elusive Majorana particles, the chip enables digital control and hardware-level error resistance. Why it matters: It could shorten the timeline for commercial-scale quantum systems from decades to years, supercharging scientific discovery when paired with AI.
Google just unveiled AI co-scientist, a multi-agent system built on Gemini 2.0 designed to serve as a virtual scientific collaborator. Instead of just summarizing papers, it uses specialized agents to generate novel hypotheses, draft proposals, and refine ideas through test-time compute scaling. Early lab tests in drug repurposing for leukemia have already shown real-world promise.
Microsoft researchers just introduced Magma, a multimodal foundation model designed to act in both digital and physical spaces. Using visual marking for action grounding and trajectory planning, it can navigate mobile UIs and steer robotic arms to grab objects. Why it matters: it moves AI beyond isolated chat interfaces into a single unified agent that handles software and hardware alike.
OpenAI just launched SWE-Lancer, a benchmark testing whether frontier LLMs can tackle real freelance engineering work. Built from over 1,400 Upwork tasks valued at $1 million in payouts—spanning $50 bug fixes to $32,000 feature builds—it found that today's top models still fail at most tasks.
Researchers just introduced MagicArticulate, a framework that automatically turns static 3D meshes into articulation-ready assets for animation. The team also built Articulation-XL, a benchmark of over 33,000 curated 3D models, using auto-regressive transformers for skeletons and diffusion for skinning to kill manual rigging.
