Amazon Launches AI Chip; Uber AI Beats Montezuma's Revenge
AWS Announces AI Chips & 13 New ML Features; Consolidating Its Cloud Dominance
Amazon Web Services has unveiled two chips and 13 machine learning capabilities and services at its AWS re:Invent conference in Las Vegas. The releases reflect Amazon’s determination to attract more developers to AWS by broadening its range of tools and services. The stock market reacted favourably, with Amazon shares rising six percent after the announcement.
How to Get a Better GAN (Almost) for Free: Introducing the Metropolis-Hastings GAN
Uber Engineering offers an alternative idea: leveraging the discriminator to pick better samples from the generator after training is done. The main idea of this method and of Discriminator Rejection Sampling is to use information from the trained discriminator in order to choose samples from the generator that are closer to samples from the real data distribution.
Uber AI Beats Montezuma’s Revenge (Video Game)
Another video game has succumbed to the strength of artificial intelligence. Uber researchers announced yesterday that their AI has completely solved Atari's Montezuma's Revenge, a classic game that involves moving a character from one room to another while killing enemies and collecting jewels in a 16th century Aztec-like pyramid.
Alibaba Open-Sources Its X-Deep Learning Framework
Alibaba has announced it will open-source X-Deep Learning (XDL), the algorithm framework behind its marketing technology and big data platform Alimama. The source code and support documents’ release is slated for December.
Technology
Neural Egg Separation: Training ML Systems to Extract Audio/Video from Noisy/Cluttered Environments
A new method for identifying distinct images and sounds within noisy environments, a long-standing challenge for machine learning (ML) systems. Called Neural Egg Separation (NES) — a reference to separating egg whites and yolks — this approach isolates audio and visual sources through a series of comparisons between signals that are clear and ones that are obscured.
GAN Dissection: Visualizing and Understanding Generative Adversarial Networks
In this work, researchers study the internal representations of GANs. To a human observer, a well-trained GAN appears to have learned facts about the objects in the image: for example, a door can appear on a building but not on a tree. Researchers wish to understand how a GAN represents such a structure.
Physics-Based Learned Design: Teaching a Microscope How to Image
Optimal experimental design methods are useful when the system is linear; however, their design will not necessarily improve the performance when the system is non-linear. In this article, researchers have shown they can optimize the experimental design for a non-linear computational imaging system using supervised learning and an unrolling physics-based network.
You May Also Like
Michael I. Jordan Interview: Clarity of Thought on AI
Jordan believes each of these should be resolved separately and on their own merits, and that a human-imitative AI system should not be the principal strategy for approaching IA and II problems.
FAIR Paper Questions Pre-Training’s Efficacy in CV
This paper suggests that while many researchers believe the path to solving computer vision challenges is “paved by pre-training a ‘universal’ feature representation on ImageNet-like data” — the process may not actually be helpful at all.
Global AI Events
| 3–5 Dec | AI World | Boston, USA |
| 3–8 Dec | NIPS | Montréal, Canada |
| 5–6 Dec | AI Summit New York | New York, USA |
| 7 Dec | Machine Learning Innovation Summit | Dublin, Ireland |
