I/O 2019 | Your Data Stays on Your Phone: Google Promises a Better AI
The success of artificial intelligence is built on large corpuses of centralized data collection, but rising concerns over user privacy and data misuse have left many people wary of fully embracing AI on their mobile devices. Google wants to change that.
Google I/O 2019
– Google I/O 2019 Keynote
– Google I/O 2019 | Geoffrey Hinton Says Machines Can Do Anything Humans Can
– Google Adds Translation, Object Detection and Tracking, and AutoML Vision Edge to ML Kit
– Google Cloud Makes Pods with 1,000 TPU Chips Available in Public Beta
– Google Is Testing Mini-Apps in Search and Google Assistant
– Google’s Project Euphonia Wants to Make Voice Recognition Work for People with Speech Impairments
– Google Brings Augmented Reality to Search
– Google Renames Home Hub to The Nest Hub and Releases A 10-inch Nest Hub Max
Microsoft Build 2019
– Microsoft Build 2019 Keynote
– AI-Driven Collaboration and Hybrid-Cloud Innovations for Microsoft 365 & Azure
– Microsoft Is Making Cortana Better at Holding Conversations
– Microsoft Introduces Power BI Embedded Enhancements and PowerApps updates
– Microsoft Launches A New Platform for Building Autonomous Robots
Papers from ICLR 2019
– MILA, Microsoft, and MIT Share Best Paper Honours
– Best Paper of ICLR 2019 | Ordered Neurons: Integrating Tree Structures into Recurrent Neural Networks
– Best Paper of ICLR 2019 | The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks
– Tsinghua, Google and ByteDance Propose Neural Networks for Inductive Learning & Logic Reasoning
MixMatch: A Holistic Approach to Semi-Supervised Learning
In this work, researchers unify the current dominant approaches for semi-supervised learning to produce a new algorithm, MixMatch, that works by guessing low-entropy labels for data-augmented unlabeled examples and mixing labeled and unlabeled data using MixUp.
(Ian Goodfellow & Google Research)
Few-Shot Unsupervised Image-to-Image Translation
Researchers seek a few-shot, unsupervised image-to-image translation algorithm that works on previously unseen target classes that are specified, at test time, only by a few example images. Their model achieves this few-shot generation capability by coupling an adversarial training scheme with a novel network design.
(NVIDIA & Cornell University & Aalto University)
Adversarial Examples Are Not Bugs, They Are Features
Researchers demonstrate that adversarial examples can be directly attributed to the presence of non-robust features: features derived from patterns in the data distribution that are highly predictive, yet brittle and incomprehensible to humans.
You May Also Like
An Introductory Guide to Computer Vision
In this guide, you’ll learn about the basic concept of computer vision and how it’s used in the real world. It’s a simple examination of a complex problem for anybody who has ever heard of computer vision but isn’t quite sure what it’s all about and how it’s applied.
‘Abandon US’ Petition Protests AI Conference Visa Denials
A French-based research scientist has collected almost 300 signatures on an online petition calling for organizers of major computer science conferences such as SIGGRAPH, NeurIPS, ICML, CVPR, etc. to exclude the US as a host.
Global AI Events
May 23-24: Deep Learning Summit in Boston, United States
May 27-28: SingularityU Nordic Summit in Helsinki, Finland
June 4-7: Amazon re:MARS in Las Vegas, United States
June 15-21: Computer Vision and Pattern Recognition in Long Beach, United States
June 10-12:World Conference on Robotics and AI (WCRAI) in Osaka, Japan
Global AI Opportunities