|December 6 · Issue #18 · View online |
A lot of news are coming out of this year’s NIPS, Uber opens an AI lab dedicated to cutting-edge research in artificial intelligence and machine learning, DeepMind open sources DeepMind Lab a fully 3D game-like platform tailored for agent-based AI research, Universe is a similar platform by OpenAI for measuring and training an AI’s general intelligence across the world’s supply of games, websites and other applications, Apple will finally start publishing its AI research and much much more.
Head right in and enjoy!
As always we appreciate you sharing this newsletter with your friends and colleagues.
| Founding Uber AI Labs |
Uber joins the AI game by founding its own research lab. Currently consisting of the Geometric Intelligence team, an AI focused startup that was acquired by Uber.
| Amazon Go |
Amazon goes back to offline shopping and applies Deep Learning to make it easy for you. More buzzwords than details, but we’re excited about whats to come.
| Do machines actually beat doctors? |
Luke Oakden-Rayner gives an overview on why Deep Learning won’t actually beat doctors in the near future.
| Robots won’t kill the workforce. They’ll save the global economy. |
The labor pool isn’t growing fast enough to support the world’s economies. So we need to teach the robots to help us.
| Apple is finally going to start publishing its AI research |
Apple announced at NIPS that they will start publishing some of the results they make in the Deep Learning field.
| NVIDIA Launches Deep Learning Teaching Kit for University Professors |
NVIDIA continues their Deep Learning efforts and starts offering specialised kits for universities.
| Generative Adversarial Networks (GANs) |
Ian Goodfellows introduction to GANs, fresh from NIPS 2016.
| An Interactive Tutorial on Numerical Optimisation |
An awesome article that lets you explore different optimisation strategies, try different learning rates, see how the gradient descent changes!
| Deep Learning the Stock Market |
Learn about embeddings, CNNs and RNNs while trying to make predictions about the stock market. A great article and fascinating read.
| Decoding the Thought Vector |
This article tries to find out whats contained in a “thought vector” or embedding. Well written and aided by interactive visualisations, just how we like it.
| Finding the genre of a song with Deep Learning |
Ever wanted to classify your music library? Find out how to do so using a CNN while learning a lot about working with audio data in Julien Despois article. TF-Learn implementation included.
| DeepMind Lab |
DeepMind announced the release of their 3D platform tailored for agent training. It offers rich science fiction-style visuals, endless extensions and is highly customisable.
| Universe |
OpenAI made a a similar announcement with Universe. It allows an AI agent to use a computer like a human does: by looking at screen pixels and operating a virtual keyboard and mouse. And there are thousands of flash games, browser games and video games to train on.
| Procedural Generation of Videos to Train Deep Action Recognition Networks |
Training on videos is hard, because there are not many labeled video datasets. So Souza et al. developed a learning architecture to generate such videos.
| robbiebarrat/rapping-neural-network |
The title says it all. Let a neural net write your lyrics.