This work was made possible by core research and engineering efforts from Andrea Agostinelli, Zalán Borsos, George Brower, Antoine Caillon, Cătălina Cangea, Noah Constant, Michael Chang, Chris Deaner, Timo Denk, Chris Donahue, Michael Dooley, Jesse Engel, Christian Frank, Beat Gfeller, Tobenna Peter Igwe, Drew Jaegle, Matej Kastelic, Kazuya Kawakami, Pen Li, Ethan Manilow, Yotam Mann, […]
Most systems we regularly interact with, such as computer operating systems, are faced with the challenge of providing good performance, while managing limited resources like computational time and memory. Since it is challenging to optimally manage these resources, there is increasing interest in the use of machine learning (ML) to make this decision-making data driven
Examining DeepVariant To better understand what DeepVariant is learning from its training data, we used a set of simple clustering and visualization methods to summarize the information captured in the model’s high dimensional data. In partnership with collaborators on the Google Genomics team, we first loaded examples into the Integrated Genomics Viewer (IGV), a widely-used
Over recent years, developers and researchers have made progress in efficiently building AI applications. Google Research has contributed to this effort by providing easy-to-use embedding APIs for radiology, digital pathology and dermatology to help AI developers train models in these domains with less data and compute. However, these applications have been restricted to 2D imaging,
To understand the probabilistic reasoning capabilities of three state-of-the-art LLMs (Gemini, GPT family models), we define three distinct tasks: estimating percentiles, drawing samples, and calculating probabilities. These tasks reflect key aspects of interpreting probability distributions, such as understanding where a sample falls within a distribution (percentiles), generating representative data (sampling), and assessing the likelihood of
High dynamic range (HDR) photography techniques can accurately capture a scene’s full range of brightness values — from those of its darkest shadows to its brightest light sources. However, when the time comes to view the resulting HDR photos, many displays are only capable of showing a limited range of brightness levels. This discrepancy means
Noise disrupts quantum correlations, effectively shrinking the available quantum circuit volume. We seek to understand if it’s possible to harness the full quantum circuit volume of a processor despite the effect of noise. In other words, we explore if it would be possible to realize an equivalent computation on a quantum processor of a smaller
This morning, Co-founder and CEO of Google DeepMind and Isomorphic Labs Sir Demis Hassabis, and Google DeepMind Director Dr. John Jumper were co-awarded the 2024 Nobel Prize in Chemistry for their work developing AlphaFold, a groundbreaking AI system that predicts the 3D structure of proteins from their amino acid sequences. David Baker was also co-awarded
The implementation of XR-Objects involves four steps: (1) detecting objects, (2) localizing and anchoring onto objects, (3) coupling each object with an MLLM for metadata retrieval, and (4) executing actions and displaying the output in response to user input. We use Unity and its AR Foundation to bring these together to build a system that
