Learning human objectives by evaluating hypothetical behaviours

We present a new method for training reinforcement learning agents from human feedback in the presence of unknown unsafe states.Read More

Related Google News:

Databricks on Google Cloud: an open integrated platform for data, analytics and machine learning February 17, 2021
Uncovering Unknown Unknowns in Machine Learning February 11, 2021
Can machine learning make you a better athlete? February 4, 2021
Machine Learning for Computer Architecture February 4, 2021
Evaluating Design Trade-offs in Visual Model-Based Reinforcement Learning February 3, 2021
Learning to Reason Over Tables from Less Data January 29, 2021
Using machine learning to improve road maintenance January 13, 2021
How to automatically scale your machine learning predictions December 17, 2020

December 13, 2019

AI / Google

Google AIFollow

Google AI@GoogleAI·

25 Feb

Lyra is a new high-quality, very low-bitrate speech codec that leverages advances in #MachineLearning to make voice communication possible even on the slowest networks. Read all about it below: http://goo.gle/3ksSSLl

Reply on Twitter 1365047309990387713 Retweet on Twitter 1365047309990387713191 Like on Twitter 1365047309990387713710 Twitter 1365047309990387713

Google AI@GoogleAI·

23 Feb

Today we present the technology behind the new Cinematic photos feature in Google Photos, where #MachineLearning is used to add simulated camera motion and parallax to a single-frame still photo, transforming a 2D photo into a more immersive 3D scene. https://goo.gle/3aKeRKr

Reply on Twitter 1364304272519663625 Retweet on Twitter 1364304272519663625195 Like on Twitter 1364304272519663625834 Twitter 1364304272519663625

Google AI@GoogleAI·

19 Feb

We are excited to announce the release of Model Search, an open-source, domain-agnostic #AutoML platform that helps researchers to efficiently and automatically develop optimal #MachineLearning models. Learn more at http://goo.gle/3dtl2o5

Reply on Twitter 1362841864622104576 Retweet on Twitter 1362841864622104576480 Like on Twitter 13628418646221045761383 Twitter 1362841864622104576

Google AI@GoogleAI·

18 Feb

Presenting DreamerV2, the first world model-based #ReinforcementLearning agent to achieve top-level performance on the Atari benchmark, learning general representations from images to discover successful behaviors in latent space. Read more at http://goo.gle/3ub8iZe

Reply on Twitter 1362481603759407105 Retweet on Twitter 1362481603759407105274 Like on Twitter 13624816037594071051042 Twitter 1362481603759407105

Google AI@GoogleAI·

16 Feb

Can models more efficiently learn rearrangement tasks by overlaying 3D space instead of using object-centric representations? Check out Transporter Nets, an open-source framework for sample-efficient robot manipulation, with related benchmark tasks. See ↓ http://goo.gle/37k9KOW

Reply on Twitter 1361755599994724353 Retweet on Twitter 1361755599994724353147 Like on Twitter 1361755599994724353602 Twitter 1361755599994724353

Google AI@GoogleAI·

11 Feb

Announcing the release of TensorFlow 3D, a set of training and evaluation pipelines for state-of-the-art 3D semantic segmentation, object detection and instance segmentation, with support for distributed training. Check it out and download the code at http://goo.gle/3pchcSG

Reply on Twitter 1359988849045164032 Retweet on Twitter 1359988849045164032343 Like on Twitter 13599888490451640321143 Twitter 1359988849045164032

Google AI@GoogleAI·

11 Feb

Introducing the CATS4ML Data Challenge, a competition to identify “unknown unknowns”, data classes that are difficult for #ML models to evaluate due to the lack of real-world examples in training data. Learn more and join the challenge ↓ http://goo.gle/3tQvWKs

Reply on Twitter 1359951245075566592 Retweet on Twitter 135995124507556659288 Like on Twitter 1359951245075566592279 Twitter 1359951245075566592

Google AI@GoogleAI·

5 Feb

Introducing TracIn, a simple, scalable approach to understand the influence of training data on #MachineLearning models by tracing the changes in prediction as individual training examples are visited during the training process. Learn all about it below: https://goo.gle/3jkclgT

Reply on Twitter 1357766939855048704 Retweet on Twitter 1357766939855048704231 Like on Twitter 1357766939855048704643 Twitter 1357766939855048704

Google AI@GoogleAI·

4 Feb

Check out new work on ML-driven design and exploration of custom accelerators, showing how #MachineLearning facilitates architecture exploration by rapidly identifying high-performing configurations across a range of applications. Learn more ↓ http://goo.gle/3rngiEh

Reply on Twitter 1357398464863166465 Retweet on Twitter 1357398464863166465140 Like on Twitter 1357398464863166465398 Twitter 1357398464863166465

Google AI@GoogleAI·

3 Feb

Introducing the World Models Library, an open-source, platform-agnostic suite of tasks and tools for examination of world model design and performance in visual model-based reinforcement learning. Learn more and grab the code at http://goo.gle/36EXY1t

Reply on Twitter 1357049407632330752 Retweet on Twitter 1357049407632330752326 Like on Twitter 13570494076323307521188 Twitter 1357049407632330752