We’re kicking off 2023 with a series of blog posts that look back at some of the many new and exciting developments coming out of Google Research. Check out our first post from @JeffDean
, about computer vision, language, multimodal, and generative models! https://
goo.gle/3CWSrU1
@googleai
-
Google Research 2023: Computer Vision, Language and Generative Models
By
–
-
EHR-Safe Generates Privacy-Preserving Synthetic Electronic Health Records
By
–
Analyzing electronic health records (EHRs) has great potential, e.g. to enhance patient care, but common anonymization methods can decrease the data’s utility. To that end, read how EHR-Safe generates high-fidelity & privacy-preserving synthetic EHR data→ https://
goo.gle/3HZfpxn -
Project Relate: Machine Learning Speech Recognition for Non-Standard Speakers
By
–
Following years of research featuring thousands of people who recorded over a million speech samples, we released Project Relate in beta. This Android app uses machine learning to offer personalized speech recognition for non-standard speakers. https://
g.co/projectrelate -
Connect-the-Dots: New Algorithm for Differential Privacy Loss Analysis
By
–
Introducing Connect-the-Dots, a new accounting algorithm that uses an indirect approach to accurately discretize privacy loss distributions, yielding a useful tool for understanding the privacy cost of combinations of differential privacy algorithms. https://
goo.gle/3FIoFmD -
Award-Winning Research on Language Grounding in Robotic Affordances
By
–
Congratulations to the authors of “Do As I Can, Not As I Say: Grounding Language in Robotic Affordances” (see the blog at https://
goo.gle/3QRJhgl) for winning the Special Innovation Award at @corl_conf
! #CoRL2022 -
CALM: Dynamic Computational Effort for Language Models
By
–
Presenting Confident Adaptive Language Modeling (CALM), a novel method that allows language models to dynamically modify computational effort when generating text. Learn how CALM can accelerate text generation while preserving output quality → https://t.co/Nm7yyT8sMA pic.twitter.com/MpuaPBzDMU
— Google AI (@GoogleAI) 16 décembre 2022Presenting Confident Adaptive Language Modeling (CALM), a novel method that allows language models to dynamically modify computational effort when generating text. Learn how CALM can accelerate text generation while preserving output quality → https://
goo.gle/3HJKzbM -
Google Recorder App Introduces Speaker Labels with On-Device AI
By
–
Learn more about speaker labels for the Recorder app, an opt-in feature based on a novel speaker diarization system for streaming on-device applications that annotates recording transcripts in real time with unique and anonymous labels for each speaker → https://t.co/TQHnZlQ3XS pic.twitter.com/0qpD2hx5YQ
— Google AI (@GoogleAI) 14 décembre 2022Learn more about speaker labels for the Recorder app, an opt-in feature based on a novel speaker diarization system for streaming on-device applications that annotates recording transcripts in real time with unique and anonymous labels for each speaker → https://
goo.gle/3HCN5k8 -
Robotics Transformer 1: Multi-Task Robot Learning Model
By
–
Introducing the Robotics Transformer 1, a multi-task model that tokenizes robot inputs and outputs actions to enable efficient inference at runtime. Learn how it improves zero-shot generalization to new tasks, environments and objects → https://t.co/hnsKvCJjmP pic.twitter.com/g9QFXzjs9T
— Google AI (@GoogleAI) 13 décembre 2022Introducing the Robotics Transformer 1, a multi-task model that tokenizes robot inputs and outputs actions to enable efficient inference at runtime. Learn how it improves zero-shot generalization to new tasks, environments and objects → https://
goo.gle/3Yxomnt -
Google Unveils Learning Interpretability Tool and Salience Evaluation Protocol
By
–
Today at 12:30pm the Google booth will host a demo about Google's open source Learning Interpretability Tool (LIT; https://
goo.gle/3kTIiLT) along with a newly introduced evaluation protocol (
https://
goo.gle/3gZSOWl) that illustrates the difference between various salience methods -
T-STAR: Style Transfer with AMR Graphs for Content Preservation
By
–
T-STAR is a style transfer approach that uses AMR graphs for intermediate representations. The first of it's kind, T-STAR yields high content preservation with negligible accuracy loss. Drop by the @emnlpmeeting Google booth at 3:30pm today to hear @JangraAnubhav talk about it!