Google AIVerified account

@GoogleAI

Google AI is focused on bringing the benefits of AI to everyone. In conducting and applying our research, we advance the state-of-the-art in many domains.

Mountain View, CA
research.google
Joined April 2009
349 Photos and videos Photos and videos

Tweets

You blocked @GoogleAI

Are you sure you want to view these Tweets? Viewing Tweets won't unblock @GoogleAI

  1. Sep 25

    Presenting the results of the Instance Level Recognition Workshop from , including a discussion of the new DELG model for ILR, new resources and open-sourced code bases, and two challenges for landmark recognition and retrieval tasks. More →

    1 reply 63 retweets 206 likes
    Undo
  2. Sep 24

    Introducing TensorFlow Recommenders, an open-source package that makes building, evaluating, and serving sophisticated recommender models easy

    ☝ We have a recommendation for that Introducing TensorFlow Recommenders, an open-source package that makes building, evaluating, and serving recommender models easy. Find recommendations for movies, restaurants, and much more! Get started →
    1 reply 170 retweets 624 likes
    Undo
  3. Sep 21

    Today we describe a model that achieves near BERT-level performance on text classification tasks, while using orders of magnitude fewer model parameters. Learn all about it below:

    10 replies 298 retweets 1,020 likes
    Undo
  4. Sep 18

    DeepVariant 1.0, the latest release of an open-source tool for isolating genome variants from sequencing data, significantly reduces sequencing errors to achieve Best Overall accuracy in 3 categories of the v2 Truth Challenge. Read more ↓

    4 replies 116 retweets 309 likes
    Undo
  5. Sep 17

    We are committed to helping researchers spend more of their time understanding instead of wrangling data, while keeping user privacy and security at the forefront.

    To support the global community in its fight against , we've collaborated with , & to launch datasets & modeling tools that are accessible to the public to help address this health & economic crisis.
    Show this thread
    5 replies 45 retweets 128 likes
    Undo
  6. Sep 16

    RigL is a new algorithm for training sparse neural networks. Instead of pruning a pre-existing dense network, it dynamically builds one during training without sacrificing accuracy relative to traditional approaches. Learn how it’s done at

    14 replies 387 retweets 1,279 likes
    Undo
  7. Sep 15

    Introducing an approach for the low-data regime that calculates the Wasserstein distance (aka the earth mover’s distance) between expert and agent, yielding excellent performance with limited demonstrations or environment interaction.

    8 replies 130 retweets 492 likes
    Undo
  8. Sep 3

    Today we present the technology behind our recently expanded flood forecasting system, which includes new morphological inundation models and alert-targeting models, plus a look at our next generation water level model, HydroNets. Read more at

    14 replies 214 retweets 621 likes
    Undo
  9. Sep 2

    Introducing KeyPose, a new model that estimates 3D keypoints of transparent objects from monocular or stereo images without explicitly computing depth, demonstrating state-of-the-art results for depth estimation. Learn all about it at

    11 replies 204 retweets 630 likes
    Undo
  10. Sep 2

    We are proud to partner with , a leader in advancing science & technology, to support a National AI Research Institute focused on Human-AI Interaction and Collaboration. To learn more and find the NSF solicitation, visit our post on the Keyword at ↓

    7 replies 101 retweets 432 likes
    Undo
  11. Aug 28

    Announcing C2D2, a -based approach to improving colonoscopy screening coverage that performs real-time local 3D reconstruction of the colon during the procedure, and identifies regions outside the field of view of the endoscope. Read more ↓

    7 replies 162 retweets 428 likes
    Undo
  12. Aug 27

    In “Hartree-Fock on a Superconducting Qubit Quantum Computer”, appearing today in , the Google AI Quantum team highlights details behind the largest chemical simulation performed on a quantum computer to date. Learn more and read the paper→

    23 replies 300 retweets 979 likes
    Undo
  13. Aug 26

    Axial-DeepLab is a novel network for image segmentation that enables long-range attention by replacing all convolution layers with sequential 1D height and width self-attention modules, resulting in state-of-the-art performance. Learn more below:

    8 replies 200 retweets 736 likes
    Undo
  14. Aug 25

    A new study of the >31M datasets cataloged by analyzes their accessibility and usability, makes recommendations for dataset publishers, and includes a metadata dataset release to encourage additional research. Read more and get the data ↓

    3 replies 180 retweets 501 likes
    Undo
  15. Aug 24

    And another shoutout to and co-authors, who received the Koenderink Prize for their 2010 paper, "Improving the Fisher Kernel for Large-Scale Image Classification" ()!

    8 retweets 33 likes
    Show this thread
    Show this thread
    Undo
  16. Aug 24

    We're proud to be a Platinum Partner of the 16th . Special shout-out to Googler Jon Barron and co-authors for their Honorable Mention Award for "NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis". See more papers ↓

    2 replies 39 retweets 168 likes
    Show this thread
    Show this thread
    Undo
  17. Aug 21

    In an effort to better understand how view selection affects models, we analyze the impact of viewpoint mutual information on downstream task performance. Learn more and grab the supporting code with pre-trained models at:

    3 replies 118 retweets 470 likes
    Undo
  18. Aug 20

    To address the challenges unique to offline , we are releasing an open-source benchmark, D4RL, as well as a simple and effective offline RL algorithm, called conservative Q-learning (CQL). Read all about it at

    3 replies 164 retweets 433 likes
    Undo
  19. Aug 19

    To better understand the impact of noisy labels on model training, we are announcing MentorMix, a new method to mitigate the impact of noisy labels, as well as a benchmark and dataset on real-world label noise. Learn more about it at:

    5 replies 194 retweets 507 likes
    Undo
  20. Aug 18

    Introducing LaBSE, a multilingual BERT model for the generation of cross-lingual sentence embeddings that exhibits exceptional performance for both high- and low-resource languages. Learn more, including how to get the pre-trained model, at

    8 replies 240 retweets 678 likes
    Undo

@GoogleAI hasn't Tweeted yet.

Loading seems to be taking a while.

Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.

    You may also like

    ·