|July 10 · Issue #89 · View online |
Hey and Welcome to a new week in deep learning!
As always, we hope you’ll enjoy reading as much as we did and would appreciate you sharing this newsletter with friends and colleagues.
See you next week!
| Bias Detectives: The Researchers Striving to Make Algorithms Fair |
An interesting and measured article, exploring the complex questions about what it means to make an algorithm fair and that detecting bias in machine learning algorithms might not be as straightforward as one might expect.
| Facial recognition software is not ready for use by law enforcement |
This article explores the use of facial recognition for things like the social credit system currently being built in China and why America and the technology still have a long way to go before being usable in anything that critical.
| Learning to drive in a day |
A fun to read and fun to watch piece on training a neural network to drive a car. The catch is, it’s not a virtual car in a 3D environment, but a real car on a real road instead. Trained with nothing more than a little time and some corrections from its human driver. Very impressive.
| Nvidia uses artificial intelligence to fake realistic slow-motion video |
Interesting new work from Nvidia, that allows further slowing down slow motion videos. Although rather limited to specific use cases, the idea of generating intermediate images using a neural network makes a lot of sense.
| Join Company Builder Entrepreneur First (Sponsored) |
Entrepreneur First funds exceptional individuals to build deep tech AI and ML companies, pre-team and pre-idea. Join our programme in London, Paris, Berlin, Singapore or Hong Kong. Find out more at www.joinef.com
| A New Angle on L2 Regularization |
Very extensive exploration of the way L2 regularization influences neural networks and how it’s connected to adversarial examples. Very nice visualizations and explanations, we definitely recommend reading this one!
| Recurrent Neural Networks: The Powerhouse of Language Modeling |
This is a nice introduction to recurrent neural networks, how they work and why they are so powerful for tasks like natural language understanding.
| My AI-Art Talk for Resonate |
Ever wanted to create art using AI? In this blog post, you’ll learn about three fun little creations and how they were made.
| Amazon.com sales rank data for Kindle and print books (61,000 books, 200,000,000 data points) |
An interesting dataset collected by NovelRank
offering detailed sales data for a set of more than 60k books over the past one and a half years.
| Patterns for Fast Prototyping with TensorFlow |
A rather short post, but it contains valuable tips on how to use some more hidden functionality of Python to easier build your models. Especially avoiding accidental config modification and wrapping a graphs endpoints seem like a great idea.
| Train a model in tf.keras with Colab, and run it in the browser with TensorFlow.js |
Zaid Alyafeai shows whats possible with Googles Notebook environment, Keras and TensorFlowJS: Building and training a neural network capable of recognizing drawings and deploying it in a browser.
| How fast is my model? |
Very extensive look at the performance and memory requirements one should care about. The whole article is focused on iOS, but the assumptions and explanations hold true for any framework currently available.
| O'Reilly Artificial Intelligence Conference in San Francisco |
The AI Conference returns to San Francisco this September. No other conference combines this depth of technical expertise with a laser focus on how to apply AI in your products and in your business today. Save 20% on most passes with the code DLW20
. Last year this event sold out - don’t miss it. Register today!
| Living in an AI World: How Startups are Shaping Tomorrow |
A curated discussion on the roles of entrepreneurs and governments in the AI era regarding education, the job market, ethics, inclusion and social impact. Come, engage and be part of the conversation with policymakers, AI entrepreneurs, and venture capitalists. July 19th, at Entrepreneur First
HQ in London. Sign up now
| Capture the Flag: the emergence of complex cooperative agents |
And last but not least, following OpenAI, DeepMind has announced impressive new work in RL, which tackled Quake, a 3D multiplayer first-person video game. Using only pixels and game points as input, their agents achieve human-level while learning and acting independently to cooperate and compete with other agents. Very impressive and the blog presentation with an interactive game explorer is wonderful.
| Self-Supervised Tracking via Video Colorization |
Google researchers managed to learn object tracking without labeled data by training a neural network to colorize grayscale videos. By starting off with a single colored frame, the model learned to color all subsequent frames, which now allows tracking objects by initially coloring them. It can follow multiple objects, track through occlusions, and remain robust over deformations without requiring any labeled training data.