Mustafa Suleyman

Beatriz Costa-Gomes, Pavel Tolmachev, Eloise Taysom et al.

We analyze over 500,000 de-identified health-related conversations with Microsoft Copilot from January 2026 to characterize what people ask conversational AI about health. We develop a hierarchical intent taxonomy of 12 primary categories using privacy-preserving LLM-based classification validated against expert human annotation, and apply LLM-driven topic-clustering for prevalent themes within each intent. Using this taxonomy, we characterize the intents and topics behind health queries, identify who these queries are about, and analyze how usage varies by device and time of day. Five findings stand out. First, nearly one in five conversations involve personal symptom assessment or condition discussion, and even the dominant general information category (40%) is concentrated on specific treatments and conditions, suggesting that this is a lower bound on personal health intent. Second, one in seven of these personal health queries concern someone other than the user, such as a child, a parent, a partner, suggesting that conversational AI can be a caregiving tool, not just a personal one. Third, personal queries about symptoms and emotional health queries increase markedly in the evening and nighttime hours, when traditional healthcare is most limited. Fourth, usage diverges sharply by device: mobile concentrates on personal health concerns, while desktop is dominated by professional and academic work. Fifth, a substantial share of queries focuses on navigating healthcare systems such as finding providers, and understanding insurance, highlighting friction in the delivery of existing healthcare. These patterns have direct implications for platform-specific design, safety considerations, and the responsible development of health AI.

Stanislav Nikolov, Sam Blackwell, Alexei Zverovitch et al.

Over half a million individuals are diagnosed with head and neck cancer each year worldwide. Radiotherapy is an important curative treatment for this disease, but it requires manual time consuming delineation of radio-sensitive organs at risk (OARs). This planning process can delay treatment, while also introducing inter-operator variability with resulting downstream radiation dose differences. While auto-segmentation algorithms offer a potentially time-saving solution, the challenges in defining, quantifying and achieving expert performance remain. Adopting a deep learning approach, we demonstrate a 3D U-Net architecture that achieves expert-level performance in delineating 21 distinct head and neck OARs commonly segmented in clinical practice. The model was trained on a dataset of 663 deidentified computed tomography (CT) scans acquired in routine clinical practice and with both segmentations taken from clinical practice and segmentations created by experienced radiographers as part of this research, all in accordance with consensus OAR definitions. We demonstrate the model's clinical applicability by assessing its performance on a test set of 21 CT scans from clinical practice, each with the 21 OARs segmented by two independent experts. We also introduce surface Dice similarity coefficient (surface DSC), a new metric for the comparison of organ delineation, to quantify deviation between OAR surface contours rather than volumes, better reflecting the clinical task of correcting errors in the automated organ segmentations. The model's generalisability is then demonstrated on two distinct open source datasets, reflecting different centres and countries to model training. With appropriate validation studies and regulatory approvals, this system could improve the efficiency, consistency, and safety of radiotherapy pathways.

Harsha Nori, Mayank Daswani, Christopher Kelly et al.

Artificial intelligence holds great promise for expanding access to expert medical knowledge and reasoning. However, most evaluations of language models rely on static vignettes and multiple-choice questions that fail to reflect the complexity and nuance of evidence-based medicine in real-world settings. In clinical practice, physicians iteratively formulate and revise diagnostic hypotheses, adapting each subsequent question and test to what they've just learned, and weigh the evolving evidence before committing to a final diagnosis. To emulate this iterative process, we introduce the Sequential Diagnosis Benchmark, which transforms 304 diagnostically challenging New England Journal of Medicine clinicopathological conference (NEJM-CPC) cases into stepwise diagnostic encounters. A physician or AI begins with a short case abstract and must iteratively request additional details from a gatekeeper model that reveals findings only when explicitly queried. Performance is assessed not just by diagnostic accuracy but also by the cost of physician visits and tests performed. We also present the MAI Diagnostic Orchestrator (MAI-DxO), a model-agnostic orchestrator that simulates a panel of physicians, proposes likely differential diagnoses and strategically selects high-value, cost-effective tests. When paired with OpenAI's o3 model, MAI-DxO achieves 80% diagnostic accuracy--four times higher than the 20% average of generalist physicians. MAI-DxO also reduces diagnostic costs by 20% compared to physicians, and 70% compared to off-the-shelf o3. When configured for maximum accuracy, MAI-DxO achieves 85.5% accuracy. These performance gains with MAI-DxO generalize across models from the OpenAI, Gemini, Claude, Grok, DeepSeek, and Llama families. We highlight how AI systems, when guided to think iteratively and act judiciously, can advance diagnostic precision and cost-effectiveness in clinical care.

Will Kay, Joao Carreira, Karen Simonyan et al.

We describe the DeepMind Kinetics human action video dataset. The dataset contains 400 human action classes, with at least 400 video clips for each action. Each clip lasts around 10s and is taken from a different YouTube video. The actions are human focussed and cover a broad range of classes including human-object interactions such as playing instruments, as well as human-human interactions such as shaking hands. We describe the statistics of the dataset, how it was collected, and give some baseline performance figures for neural network architectures trained and tested for human action classification on this dataset. We also carry out a preliminary analysis of whether imbalance in the dataset leads to bias in the classifiers.

Arun Nair, Praveen Srinivasan, Sam Blackwell et al.

We present the first massively distributed architecture for deep reinforcement learning. This architecture uses four main components: parallel actors that generate new behaviour; parallel learners that are trained from stored experience; a distributed neural network to represent the value function or behaviour policy; and a distributed store of experience. We used our architecture to implement the Deep Q-Network algorithm (DQN). Our distributed algorithm was applied to 49 games from Atari 2600 games from the Arcade Learning Environment, using identical hyperparameters. Our performance surpassed non-distributed DQN in 41 of the 49 games and also reduced the wall-time required to achieve these results by an order of magnitude on most games.

Karl Moritz Hermann, Tomáš Kočiský, Edward Grefenstette et al.

Teaching machines to read natural language documents remains an elusive challenge. Machine reading systems can be tested on their ability to answer questions posed on the contents of documents that they have seen, but until now large scale training and test datasets have been missing for this type of evaluation. In this work we define a new methodology that resolves this bottleneck and provides large scale supervised reading comprehension data. This allows us to develop a class of attention based deep neural networks that learn to read real documents and answer complex questions with minimal prior knowledge of language structure.

Edward Grefenstette, Karl Moritz Hermann, Mustafa Suleyman et al.

Recently, strong results have been demonstrated by Deep Recurrent Neural Networks on natural language transduction problems. In this paper we explore the representational power of these models using synthetic grammars designed to exhibit phenomena similar to those found in real transduction problems such as machine translation. These experiments lead us to propose new memory-based recurrent networks that implement continuously differentiable analogues of traditional data structures such as Stacks, Queues, and DeQues. We show that these architectures exhibit superior generalisation performance to Deep RNNs and are often able to learn the underlying generating algorithms in our transduction experiments.