DeepMind has published a new paper on AlphaZero, a generalization of AlphaGo, showing that it convincingly defeats world-champion programs in both Chess and Shogi within only 24 hours of training (on a lot of TPUs). See these Reddit and HN threads for interesting discussions.

According to Nvidia, the Titan V is “the most powerful PC GPU ever created”. Nvidia claims up to 110 teraflops of performance, with 12GB of HBM2 memory, 5120 CUDA cores, and 640 “tensor cores” that are said to offer up to 9 times the deep-learning performance of its predecessor. More importantly, its color is gold!

The Facebook AI Research (FAIR) Residency Program is a one-year research training program with Facebook’s AI Research group, designed to give you hands-on experience of machine learning research. The application deadline is January 26, 2018, and a Bachelor’s degree is required.

Musk revealed the effort at a Tesla party that took place at the intelligence conference NIPS. According to Musk, “I wanted to make it clear that Tesla is serious about AI, both on the software and hardware fronts. We are developing custom AI hardware chips”. 

The NIPS conference, which took place in Long Beach this year, is the perhaps biggest gathering of AI researchers. This year’s tickets were sold out in record time, and attendance numbers skyrocketed. We expect videos from the conference to be posted online in the coming videos, with some of livestreamed keynotes already available on Facebook.

By creating user interfaces which let us work with the representations inside machine learning models, we can give people new tools for reasoning.

This post is a reply to Francois Chollet, the inventor of the Keras wrapper for the Tensorflow and Theano deep learning systems, on his essay “The impossibility of intelligence explosion.” If you liked Francois’ essay, this one is worth a read as well.

An accessible tutorial on how to implement a neural network in JavaScript with deeplearn.js. The goal is to find matches of font and background color for improved web accessibility.

The release includes performance improvements, new layers, and the ability to ship models to other frameworks (via ONNX), support for CUDA9 and CuDNNv7.

The new release adds new features to both the Unity SDK and Python API, adds new example environments, and improves the default Reinforcement Learning algorithm (PPO), addition to many bug-fixes.

Turi Create simplifies the development of custom machine learning models. You don’t have to be a machine learning expert to add recommendations, object detection, image classification, image similarity or activity classification to your app.

In this paper, the authors present single AlphaZero algorithm that can achieve, tabula rasa, superhuman performance in many challenging domains. Starting from random play, and given no domain knowledge except the game rules, AlphaZero achieved within 24 hours a superhuman level of play in the games of chess, and shogi (Japanese chess) as well as Go, and convincingly defeated a world-champion program in each case.

Indexes are models: a B-Tree-Index can be seen as a model to map a key to the position of a record within a sorted array, a Hash-Index as a model to map a key to a position of a record within an unsorted array, and a BitMap-Index as a model to indicate if a data record exists or not. In this exploratory research paper, the authors posit that all existing index structures can be replaced with other types of models, including deep-learning models, which we term learned indexes. Initial results show, that by using neural nets we are able to outperform cache-optimized B-Trees by up to 70% in speed while saving an order-of-magnitude in memory over several real-world data sets.

Learning CNN structures that is more efficient than previous approaches: Instead of using reinforcement learning (RL) or genetic algorithms (GA), the authors use a sequential model-based optimization (SMBO) strategy. The learned model achieves state-of-the-art performance on ImageNet.