Google DeepMind's Demis Hassabis proposed training a large language model only on scientific knowledge available before 1911 to see if it could independently derive general relativity, calling it a meaningful test for artificial general intelligence. Several research teams, including one led by Ido Kaminer at Technion, have since built such 'vintage' AI models restricted to historical data, but early results have exposed shortcomings in current systems rather than demonstrating breakthrough reasoning.
nature.com
· 2026-09-09
Google DeepMind has launched AlphaGenome Atlas, a publicly available database that uses its AlphaGenome AI model to predict the likely effects of every possible single-letter mutation across the human genome, covering both protein-coding and regulatory regions. The tool is free for non-commercial research use starting immediately, with a paid licensing option for commercial users, including DeepMind's own drug-discovery affiliate Isomorphic Labs, expected to follow via Google Cloud.
tech.slashdot.org
· 2026-09-08
Google DeepMind has released AlphaGenome Atlas, a 1-petabyte database that uses its AlphaGenome AI model to predict the regulatory impact of all roughly 9 billion possible single-letter mutations across the human genome. The Atlas also introduces a scoring system called AVI, which condenses complex coding and non-coding predictions into one number researchers can use to quickly identify which genetic variants deserve closer study.
blog.google
· 2026-09-08
Danijar Hafner is developing 'world models' — AI systems that simulate physical reality — and training reinforcement-learning agents inside them so robots can anticipate outcomes before acting in the real world. This model-based approach lets agents rehearse complex tasks virtually, avoiding the costly trial-and-error training typical of conventional robotics. Hafner, a former Google Brain and DeepMind researcher who worked alongside figures like Geoffrey Hinton, is now recognized as a rising talent in AI research.
technologyreview.com
· 2026-09-08
Google DeepMind has launched AlphaGenome Atlas, a database that predicts the molecular effects of every one of the roughly nine billion possible single-letter changes across the human genome. The tool is accessible via a web portal, the Antigravity development platform, and the AlphaGenome interface, and comes bundled with a new Variant Impact Score to help researchers prioritize which mutations are most likely to matter.
theverge.com
· 2026-09-08
TechCrunch has published a living glossary of AI terminology aimed at demystifying jargon used across the industry, from AGI to AI agents. The piece highlights 'opaque recurrence,' a reasoning technique reportedly used in OpenAI's new Astra model that has unsettled AI safety researchers, alongside more familiar terms like LLMs, RAG, and RLHF.
techcrunch.com
· 2026-09-07
Google DeepMind has released WeatherNext 3, an upgrade to its AI weather model that generates hourly forecasts on a 5-kilometer grid, five times finer than its predecessor. The improvement comes from feeding the model live satellite imagery and sparse weather station readings instead of relying mainly on lagging numerical prediction data.
engadget.com
· 2026-09-03
Google DeepMind has released WeatherNext 3, a new AI-driven global weather model that the company says improves precipitation forecast accuracy by up to 50% for predictions made a day or more ahead. The system is positioned as an upgrade over prior AI weather tools and traditional forecasting methods.
gizmodo.com
· 2026-09-03
Google DeepMind and Google Research launched WeatherNext 3, a deep-learning forecasting model that outperformed rival AI systems from Microsoft, Nvidia and ECMWF, as well as traditional forecasts from the US National Weather Service and ECMWF, on the Operational WeatherBench benchmark. Google says the model will soon power weather data shown in Search, Maps and Gemini, and will also be offered to researchers and developers through its cloud platform.
techcrunch.com
· 2026-09-03
Google has moved its AI-responsibility team out of the DeepMind research division as part of an internal reorganization. Researchers within the team have voiced concerns that the shift could compromise their independence and reduce their ability to identify risks in Google's newest AI models.
wsj.com
· 2026-08-26
Robotics startup Generalist has raised nearly $200 million more in an extension of its Series B round led by 8VC, pushing its valuation to $3 billion and total round funding to $600 million. The company, founded in 2024 by ex-Google DeepMind and Boston Dynamics engineers, builds AI foundation models meant to control multiple types of robots, with its Gen 1.5 model reportedly learning new tasks from short video clips.
techcrunch.com
· 2026-08-26
GoKawiil