ML4Sci #24: Massive Open Catalyst Project by Facebook and Carnegie Mellon; Trends from ICLR

Also, Google's AI moat and the Department of Justice acts on anti-trust

Hi, I’m Charles Yang and I’m sharing (roughly) weekly issues about applications of artificial intelligence and machine learning to problems of interest for scientists and engineers.

If you enjoy reading ML4Sci, send us a ❤️. Or forward it to someone who you think might enjoy it!

Share ML4Sci

As COVID-19 continues to spread, let’s all do our part to help protect those who are most vulnerable to this epidemic. Wash your hands frequently (maybe after reading this?), wear a mask, check in on someone (potentially virtually), and continue to practice social distancing.


There was so much news these past few weeks, I’ve decided to just send out an extended “In the News” section. Stay tuned - next issue will be a special collaborative issue on Batteries x AI!

📰In the News

ML

Towards ML Engineering: A Brief History Of TensorFlow Extended (TFX). A nice retrospective by the Google team for TensorFlow Extended on how they are rewriting the rules of “software engineering” for “ML engineering”. For the interested, here are some more DevOps advice from Google: “Rules of ML”. Keep an eye on this field: we’re going to see an explosion of new companies and new techniques in the field of ML DevOps

On the tail of DevOps and Google: “How AI is powering a more helpful Google”(A quick timeline). Kind of insane how Google’s AI expertise+scale of data is creating incredible products with awesome value add to customers in such a short period of time (and to the detriment of any competition).

🏥NVIDIA details their federated learning experiment to forecast COVID-19 patient oxygen needs across 20 hospitals without sharing patient data

Honestly, I’m just an aggregator of aggregators at this point. Here’s a nice reddit post that provides a nice list of resources on Graph Neural Networks

🤗NLP startup Hugging Face releases Inference API pricing - the start of commercial AI-as-a-Service

Landing AI Unveils AI Visual Inspection Platform to Improve Quality and Reduce Costs for Manufacturers Worldwide

📈Trends from ICLR submissions - nice meta review of where the cutting-edge of ML is going

Science

Lawrence Berkeley Lab develops self-guided algorithm to autonomously collect data on neutron scattering, with improved fidelity and reduced measuring time (btw, it wasn’t with deep learning). LBL also has a nice website called ml4sci.lbl.gov, so it looks like I’ve got some competition 👀

Not related to ML, but….we found a room-temperature superconductor. The only catch is it stabilizes at 10^7 times atmospheric pressure[🔒Nature] [QuantaMagazine]

Facebook, Carnegie Mellon launch Open Catalyst Project, a competition dataset with 1.3M DFT calculations in the hopes of finding new catalysts for energy storage[FB Blog] See below for figure from the dataset paper

🐦Another twitter thread (honestly, I should just become a twitter thread compiler): emerging trends in photovoltaics

Citrine, a ML+materials startup, details their DOE-funded work on determining uncertainty in DFT estimates. A good measure of a field’s maturity is the amount of research done by commercial ventures - it seems the computational materials field is definitely maturing in that area!

Physics Meets ML seminar: “Science is a verb: adopting the scientific method and best practices in AI research” by Michela Paganini@Facebook AI. Really important topic given the huge amount of messy, empirical work being done in AI

🔒Nature Review Materials: AI for 3D printing - using AI for difficult control problems. Silicon Valley enthusiasts want self-driving cars, but for me, I’m much more excited about self-driving printers.

The Science of Science

A great read about the ARPA innovation funding model

🎉Nature journals announce first open-access agreement with Max Planck Digital Library. Publishing cost to author is still exorbitant, but the tide is slowly turning toward open-access

🌎Out in the World of Tech

🚗Waymo opens fully driverless ridesharing to Phoenix general public. Another trend (self-driving cars) accelerated by COVID-19

Twitter and Facebook Contend With Concerns Over Election Interference, Censorship [🔒WSJ][NPR][Ben Thompson’s Stratechery]

“Google AI Tech will be used for virtual border wall, CBP contract shows”[The Intercept]

Policy and Regulation

⚖️The Department of Justice has finally filed its antitrust lawsuit against Google. For an in-depth analysis, you can check out Ben Thompson’s Stratechery article on US vs. Google

Thanks for Reading!

I hope you’re as excited as I am about the future of machine learning for solving exciting problems in science. You can find the archive of all past issues here and click here to subscribe to the newsletter.

Have any questions, feedback, or suggestions for articles? Contact me at ml4science@gmail.com or on Twitter @charlesxjyang