Google DeepMind continues to tackle some of the most intricate and engaging challenges in Artificial Intelligence, with a recent push revealing cutting-edge research projects, major breakthroughs, and a variety of open roles for researchers. The lab’s latest research news spans projects such as DolphinGemma, a large language model developed to decode dolphin communication, demonstrating the organization´s commitment to applying Artificial Intelligence in novel scientific domains. Other highlighted work includes advancements in adaptive Artificial Intelligence agents, improvements in three-dimensional scene creation, and innovative approaches to large language model training—all aimed at ensuring safer and smarter Artificial Intelligence systems.
Recent publications from Google DeepMind reflect both theoretical and practical progress across a spectrum of Artificial Intelligence fields. Notable works include ´Bridging Algorithmic Information Theory and Machine Learning, Part II´ by Marcus Hutter, which explores unsupervised learning and kernel methods, and ´Generative Ghosts: Anticipating Benefits and Risks of AI Afterlives´ appearing at ACM CHI 2025, indicative of the team´s engagement with societal implications. Additional significant research covers robust test mutators for kernel fuzzing, foundation models for personalized sensor creation, and novel pretraining strategies in multimodal frameworks. The momentum continues through foundational contributions such as KiVA for testing multimodal models, techniques for handling non-stationary stochastic bandits, and studies on embedding generalization.
Among recent breakthroughs, the Genie 2 world model enables unlimited, diverse simulation environments to train versatile agents. GenCast pushes the boundaries of weather prediction, using Artificial Intelligence to anticipate extreme conditions with unprecedented accuracy, while AlphaQubit advances error diagnosis within quantum computers, enhancing system reliability. To support its ambitious goals, DeepMind is actively hiring for roles including Senior Research Scientist in Information Quality, Robotics Systems Safety Engineer, and Research Scientist positions focused on multilinguality and robotics. Each career opportunity aims to attract talent eager to shape the future direction of Artificial Intelligence-powered products and scientific discovery, further cementing DeepMind’s role at the frontier of Artificial Intelligence research.
