A monthly roundup of news about Artificial Intelligence, Machine Learning and Data Science. This is an eclectic collection of interesting blog posts, software announcements and data applications from Microsoft and elsewhere that I've noted over the past month or so.
Open Source AI, ML & Data Science News
Facebook open-sources PyTorch-BigGraph for producing embeddings for graphs where the model is too large to fit in memory. Also published: an embeddings graph for 50 million Wikipedia concepts.
Visual Studio Code expands Python support, including a new variable explorer and data viewer, improved debugging capabilities, and real-time collaboration via Live Share.
Pyright, a static type-checker for Python, available as a command-line tool and a VS Code extension.
Frequently Asked Questions about Tensorflow 2.0, now in alpha testing.
VoTT, an open source annotation and labeling tool for image and video assets, has been updated to version 2.
Industry News
Google announces AI Platform (beta), an integrated platform for developers to create AI capabilities and deploy them to GCP or on premises.
New enterprise AI solutions from Google (in beta): Document Understanding AI, Contact Center AI, and Recommendations AI for retail.
Google made a number of other of AI-related announcements at Google Next, including new ML capabilities in BigQuery, AutoML Tables to generate an ML model predicting a data column with one click, updated pre-trained models for vision and natural language services, general availability of Cloud TPU v3, and more.
NVIDIA Tesla G4 GPUs are now available in Google Colab for faster training with larger models.
Microsoft News
Microsoft commits to host the world's leading environmental data sets on Azure as part of a broader sustainability initiative.
Microsoft's research in Machine Teaching: the process of augmenting models with human input rather than using data alone.
Anomaly Detector, a new Azure Cognitive Service to detect unusual values in time series data, is now in preview.
Azure Custom Vision, the transfer learning service for image classification with user-provided labels, is now generally available.
Azure Video Indexer can now be trained to recognize specific people in video from user-provided photographs.
Azure Cognitive Services are now accessible in Apache Spark workloads, via Spark ML pipelines.
You can now deploy Tensorflow models to Azure Data Box Edge with Azure ML Service and ONNX, for local inference on high-performance FPGA chips.
Azure HDInsight now supports Apache Hadoop 3.0, and updated versions of Hive, HBase, and Phoenix.
Managed MLflow is now generally available in Azure Databricks, to track experiments and manage models and projects. Microsoft has also joined the open-source MLflow project.
PowerBI now offers AutoML, to enable business analysts to build machine learning models from data without coding.
Learning resources
Advanced Natural Language Processing with spaCy, a free online course from Ines Montani, a core developer of the Python package. These tutorial notebooks on sentiment analysis by Ben Trevett also make use of spaCy.
The Economist reworks some of their worst published charts, providing a lesson in better data visualization.
Discriminating Systems: Gender, Race and Power in AI: Results from a year-long study on diversity in the AI sector from the AI Now Institute.
A tutorial on creating a serverless HTTP endpoint with Python and Azure Functions using Visual Studio Code.
Tutorial: incrporating predictions from an R model in Power BI report.
AWS Machine Learning University: free access to the curriculum used to teach Amazon staff about ML on AWS.
An introduction to reinforcement learning, with AWS RoboMaker.
Applications
Ganvatar: an impressive demonstration of synthesizing faces along semantic vector axes (age, gender, happiness).
Using the time series and outlier detection features of the Kusto Query Language to detect cyber threats with Azure Sentinel.
Forecasting orange juice sales in a grocery chain (Jupyter Notebook), using automated machine learning in Azure ML Service.
Seek, a smartphone app that uses computer vision to identify plant and animal species in real time.
Find previous editions of the monthly AI roundup here.
