Google Research

Google Research is a research organisation inside the American company Google. It is not a separate company but a division of Google. Google says the company itself began with a research paper on a search engine, published in 1998. Google Research's official site does not give a founding year for the organisation itself. Peter Norvig, well known as the co-author of an AI textbook, joined Google in 2001 and from 2005 served as its Director of Research, building up teams in machine translation, speech recognition and computer vision. Its home base is Mountain View, California, and as of October 2026 it is led by Google vice president Yossi Matias.

Google Research works on the foundations of machine learning and algorithms, computing systems and quantum computing, a range of sciences, and the use of AI for social issues such as health, climate and education. One of its best-known results is the Transformer, presented in the 2017 paper ‘Attention Is All You Need’. It is a neural network (an AI structure loosely modelled on connections in the brain) built only on ‘attention’, which calculates how strongly the words in a sentence relate to one another, and it became the basis of today's large language models. The work was introduced on the Google Research blog, and its authors included Llion Jones, then at Google Research. In April 2023, the Brain team from Google Research merged with DeepMind to form Google DeepMind.

In December 2025, Google Research introduced Titans, an architecture that lets an AI store new information in long-term memory even while it is running, and MIRAS, a theoretical framework behind it. That year it also released, with Google DeepMind and Yale University, C2S-Scale, a model with 27 billion parameters (the adjustable values inside a model) for analysing cells one by one. Google says WeatherNext, a weather-forecasting AI built with Google DeepMind, predicted the rapid strengthening of Hurricane Melissa and its landfall in Jamaica five days ahead in October 2025 (company-reported). In May 2026, papers on the research-assistant AIs ‘ERA’ and ‘Co-Scientist’ were published in the journal Nature, and ‘Gemini for Science’, a set of science research tools built on them, was announced.