Jun 17, 2017

Intuition Machine - Issue #27

By Intuition Machine • Issue #27 • View online

Welcome to the weekly newsletter covering Deep Learning Patterns, Methodology and Strategy. We’ve come up away to organize the topics to appeal to the broadest of audiences. The more general topics always are at the top, while more specialized ones are towards the bottom. We hope that this newsletter we appeal to all those interested in Deep Learning developments.

Guide to working in AI policy and strategy - 80,000 Hours

Written by Miles Brundage a researcher who works on AI policy at the University of Oxford’s Future of Humanity Institute, with help from the 80,000 Hours team and others acknowledged below.

80000hours.org • Share

How AI Can Keep Accelerating After Moore’s Law - MIT Technology Review

New ideas in chip design look likely to keep software getting smarter.

www.technologyreview.com • Share

Why Apple is struggling to become an artificial-intelligence powerhouse - The Washington Post

What you need for artificial intelligence goes against the grain of Apple’s secrecy.

www.washingtonpost.com • Share

US intelligence agencies are beginning to build AI spies — Quartz

A US intelligence director says a lot of espionage is more boring than you might think, and much of it could be handed over to artificial intelligence. “A significant chunk of the time, I will send [my employees] to a dark room to look at TV monitors to do national security essential work,”

qz.com • Share

Why is the language of transhumanists and religion so similar? | Aeon Essays

The most avid believers in artificial intelligence are aggressively secular – yet their language is eerily religious. Why?

aeon.co • Share

Can We Quantify Machine Consciousness? - IEEE Spectrum

Artificial intelligence might endow some computers with self-awareness. Here’s how we’d know

spectrum.ieee.org • Share

A (computational) linguistic farce in three acts

I am VP and Engineering Fellow at Google, where I lead work on natural-language understanding and machine learning.

www.earningmyturns.org • Share

Vertical AI Startups: Solving Industry-specific Problems by Combining AI and Subject Matter Expertise — Bradford Cross

While most of the machine learning talent works in big tech companies, massive and timely problems are lurking in every major industry outside tech.

www.bradfordcross.com • Share

Deal or no deal? Training AI bots to negotiate | Engineering Blog | Facebook Code | Facebook

Today, researchers at Facebook Artificial Intelligence Research (FAIR) have open-sourced code and published research introducing dialog agents with a new capability — the ability to negotiate.

code.facebook.com • Share

CortexNet

Inspired by the human visual system, CortexNet provides robust visual temporal representations by adding top-down feedback and lateral connections to bottom-up feed-forward connections, all of which are present in our visual cortex.

engineering.purdue.edu • Share

HRA

A major achievement in reinforcement learning research

www.maluuba.com • Share

Vicarious | Blog

Vicarious introduces the Schema Network, a generative graphical model that can simulate the future and reason about cause and effect.

www.vicarious.com • Share

Learning through human feedback | DeepMind

A central question in technical AI safety is how to tell an algorithm what we want it to do. Working with OpenAI, we demonstrate a novel system that allows a human with no technical experience to teach an AI how to perform a complex task, such as manipulating a simulated robotic arm.

deepmind.com • Share

Normalization in Deep Learning – CALCULATED CONTENT

A few days ago (Jun 2017), a 100 page on Self-Normalizing Networks appeared. An amazing piece of theoretical work, it claims to have solved the problem of building very large Feed Forward Networks (FNNs). It builds upon a Batch Normalization (BN), introduced in 2015– and is now the defacto standard for all CNNs and RNNs. But…

calculatedcontent.com • Share

Research Blog: The Machine Intelligence Behind Gboard

Most people spend a significant amount of time each day using mobile-device keyboards: composing emails, texting, engaging in social media, and more. Yet, mobile keyboards are still cumbersome to handle.

research.googleblog.com • Share

Data Loading and Processing Tutorial — PyTorch Tutorials documentation

A lot of effort in solving any machine learning problem goes in to preparing the data. PyTorch provides many tools to make data loading easy and hopefully, to make your code more readable.

pytorch.org • Share

[1706.03762] Attention Is All You Need

The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely.

arxiv.org • Share

[1706.04062] Deep Learning-Based Food Calorie Estimation Method in Dietary Assessment

We present our calorie estimation method in this paper. To estimate calorie of food, a top view and side view is needed.

arxiv.org • Share

[1706.04972] Device Placement Optimization with Reinforcement Learning

Importantly, the decision of placing parts of the neural models on devices is often made by human experts based on simple heuristics and intuitions. In this paper, we propose a method which learns to optimize device placement for TensorFlow computational graphs.

arxiv.org • Share

[1706.03922] Analyzing the Robustness of Nearest Neighbors to Adversarial Examples

In this paper, we take a step towards addressing this challenging question by introducing a new theoretical framework, analogous to bias-variance theory, which we can use to tease out the causes of vulnerability. We apply our framework to a simple classification algorithm: nearest neighbors, and analyze its robustness to adversarial examples.

arxiv.org • Share

SmoothGrad

When a machine learning model makes a prediction, often times we would like to determine which features of the input (pixels, for images) were important for the prediction.

tensorflow.github.io • Share

GitHub - uclmr/fakenewschallenge: UCL Machine Reading - FNC-1 Submission

The stance detection model submitted by the UCL Machine Reading group (UCLMR) for stage number 1 of the Fake News Challenge (FNC-1) is a single, end-to-end system consisting of lexical as well as similarity features fed through a multi-layer perceptron (MLP) with one hidden layer.

github.com • Share

GitHub - deepmind/dsprites-dataset: Dataset to assess the disentanglement properties of unsupervised learning methods

dsprites-dataset - Dataset to assess the disentanglement properties of unsupervised learning methods

github.com • Share

GitHub - xmartlabs/Bender: Easily craft fast Neural Networks on iOS! Use TensorFlow models. Metal under the hood.

Easily craft fast Neural Networks on iOS! Use TensorFlow models. Metal under the hood.

github.com • Share

Stardust: GPU-based Visualization Library

Stardust is a library for rendering information visualizations with GPU (WebGL). Stardust provides an easy-to-use and familiar API for defining marks and binding data to them. With Stardust, you can render tens of thousands of markers and animate them in real time without the hassle of managing WebGL shaders and buffers.

stardustjs.github.io • Share

Did you enjoy this issue?

By Intuition Machine

Deep Learning AI and Platformization Strategy

Tweet

Share

If you don't want these updates anymore, please unsubscribe here.

If you were forwarded this newsletter and you like it, you can subscribe here.