Founded by famous professor Andrew Ng, Landing.ai is a new Artificial Intelligence company focused on the manufacturing industry. At this point, it is still unclear what kind of products landing.ai is working on.

The Google AI China Center will have a small group of researchers supported by several hundred China-based engineers. “It will be a small team focused on advancing basic AI research in publications, academic conferences, and knowledge exchange,“ said Fei-Fei Li, the chief scientist at Google’s cloud unit who will lead the Beijing research center.

This tool provides analysis of 6,000 of the most popular opening sequences from the recent history of Go, using data from 231,000 human games and 75 games AlphaGo played against human players.

Someone used a Machine Learning algorithm to paste the face of ‘Wonder Woman’ star Gal Gadot onto an adult video. It’s not going to fool anyone who looks closely. Sometimes the face doesn’t track correctly and there’s an uncanny valley effect at play, but at a glance, it seems believable.

An excellent tutorial on the building blocks of today’s Deep Learning systems. The tutorial covers Convolutional Models, Autoregressive Models, Domain Alignment, Meta Learning, Graph Networks, and more.

A good summary of Deep Learning advancements for NLP in 2017. This post covers, pre-trained word embeddings, the sentiment neuron, SemEval 2017 results, abstractive summarization systems, unsupervised Machine Translation, and more.

Gaussian processes may not be at the center of current machine learning hype but are still used at the forefront of research – they were recently seen automatically tuning the MCTS hyperparameters for AlphaGo Zero for instance.

Several practical tips for training sequence-to-sequence models with attention, such as those used in Machine Translation, or text summarization.

MAgent is a research platform for many-agent reinforcement learning. Unlike previous research platforms that focus on reinforcement learning research with a single agent or few agents, MAgent aims at supporting reinforcement learning research that scales up from hundreds to millions of agents.

In this paper and project, the authors pose generate sound given visual input and apply learning-based methods to generate raw waveform samples given input video frames.

A detailed analysis of the ChestXray14 dataset, and why it may not be fit for training medical AI systems to do diagnostic work. Such analyses of real-world datasets are extremely important, and I hope to see more of them in the future.

The authors present Libratus, an AI that, in a 120,000-hand competition, defeated four top human specialist professionals in heads-up no-limit Texas hold’em, the leading benchmark and long-standing challenge problem in imperfect-information game solving. The game-theoretic approach uses application-independent techniques: an algorithm for computing a blueprint for the overall strategy, an algorithm that fleshes out the details of the strategy for subgames that are reached during play, and a self-improver algorithm that fixes potential weaknesses that opponents have identified in the blueprint strategy.

An approach to predict the performance of a network before training, based on its architecture. The authors develop a way to encode individual layers into vectors and bring them together to form an integrated description via LSTM. Taking advantage of the recurrent network’s expressive power, this method can reliably predict the performances of various network architectures.

Recently there has been a dramatic increase in the performance of recognition systems due to the introduction of deep architectures for representation learning and classification. However, the mathematical reasons for this success remain elusive. This tutorial will review recent work that aims to provide a mathematical justification for several properties of deep networks, such as global optimality, geometric stability, and invariance of the learned representations.