— In the News —
The part of an AI that we interact with is at the top of a very big stack of data, labor and natural resources. In this essay, the authors use an Amazon Echo as a vehicle to explore the full stack of what it takes to create an AI. This is a longread that keeps showing up in my Inbox for good reason.
Data centers use an enormous amount of energy and that's not likely to diminish anytime soon. This article in Nature takes a look at the innovative ways that data centers are reducing energy use, in spite of ever-increasing demands.
— Sponsored Link —
DataCamp is the online learning platform for data science. We help you learn Python, R, SQL and more with a hands-on approach to learning to code - short 3-4 minute videos are followed by practice challenges so you’re immediately applying what you’re learning. For a limited time you can learn more for less and make data science your superpower.
— Tools and Techniques —
Google's new What-If Tool enables users to analyze a machine learning model without writing code. Given pointers to a TensorFlow model and a dataset, the What-If Tool offers an interactive visual interface for exploring model results. The post on the Google AI blog offers a good overview. For more info and online demos, check out the What-If Tool project site >>
Michael Kaminsky's latest post offers a paradigm that provides clear separation between the work of software engineering teams and the data teams that they support. This enables each team to focus on what they do best while also removing obstacles and drudge work. The suggested paradigm won't be a surprise to people that work in large tech companies but there are good ideas here for everyone else.
There are often discussions around the web about the best tools for doing data analytics work. What's often missing in those discussions is that the work typically goes through phases and different tools may be more or less useful along the way. In this post, Roger Peng explores typical phases of an analytics project and how the tooling needs are different for each.
Jupytext is an open-source tool that can convert notebooks to and from Markdown documents, Julia, Python and R scripts. It enables a variety of use-cases, such as writing notebooks as plain text, versioning with meaningful diffs, code refactoring and collaboration.
— Data Viz —
This deep dive into boxplots explains what they are, exactly, how to use them, and how to make them.
— Career —
Derived from interviews with senior data scientists and managers, here's a map of milestones and key skills that are needed along a data science career path.