Radar trends to watch: July 2020
Trends in disruptions in COVID-19 and #BlackLivesMatter, AI, programming, social media, and cloud.
The Daya Bay Antineutrino Detector (source: Energy.gov on Flickr)
News of the past month is dominated by COVID-19 and the #BlackLivesMatter protests. I’ve considered them both “disruptions” because they’re forcing us to re-evaluate how we deal with data, health, and so many other issues. Those are disruptions I can get behind.
Disruptions: COVID-19 and #BlackLivesMatter
- The state of California has done an amazing job of releasing all of its COVID data, models, and tools. Public data is the only way we’re going to understand what works against this disease.
- Some spectacular COVID-19 research uses microfluidics and human cell cultures to investigate the effect of the virus on real, lab-grown human lungs.
- Coronavirus is changing car design: Kia is working on new materials that kill the virus, along with powerful ultraviolet lights in the dome to sanitize a car. Ford is experimenting with cars that heat themselves up to a temperature (133F for 15 minutes) that will kill the virus.
- #BlackLivesMatter, along with COVID, has drawn attention to an important thread in the AI ethics discussion: it isn’t just about bias, it’s about who is harmed, and the power structures that inflict harm. A twitter thread and a short course by @timnitgebru is important for rethinking issues, along with the Data for Black Lives statement on COVID-19 data. Along these lines, a letter signed by over 600 researchers got Springer not to publish a paper about using face recognition to detect “criminal characteristics.”
- Partially as a response to the disruption to businesses in black communities after the killing of George Floyd, and partially in response to the disruption caused by COVID, there’s been an increase in downloads of apps to help find black-owned businesses.
AI
- In a letter to Congress, IBM CEO Arvind Krishna backs off on facial recognition research and opposes the use of facial recognition for mass surveillance. Amazon has implemented a one-year moratorium on police use of Rekognition, their face recognition technology. Microsoft has also joined this movement. On the other hand, Clearview AI (along with companies like Palantir) are going full speed ahead building the surveillance state.
- Keen is a new Pinterest-like service from Google, but with AI. (What else?) It’s easy to think this is another instance of Google copycatting successful social media, but the idea of a “passive search” for topics that interest you is appealing.
- Facebook released a collection of 100,000 deep fakes for developing and training deep fake detection algorithms. Unfortunately, the results of their deep fake detection challenge were not encouraging; the winner had roughly 65% accuracy.
Programming
- Low-code and no-code frameworks are getting a lot more interesting. AWS Honeycode is a low-code development system that builds cell phone apps that run on AWS. Low code isn’t new (you could argue that spreadsheets are the first “low-code” app builders), and others (including Google) have failed at it, but this time, momentum seems to be building.
- Sweetviz is a new low-code library for exploratory data analysis (EDA).
- GitHub is the latest significant software product to replace the use of “master” and “slave” with other terms, e.g., “main” or “primary. (Note: this applies to GitHub, not yet to git itself.) The first was Drupal, back in 2014; Python made the switch in 2018. We expect to see many more projects making similar changes. RedHat is the latest organization to make this change (as of June 30, 2020).
- While we’ve seen a lot of interest in how to program quantum computers, most of the programming we’ve seen so far has been very low level—the equivalent of writing assembly language. Silq is the first high-level programming language for quantum computers.
- IBM has released a homomorphic encryption toolkit. It’s an important addition to the very short list of Homomorphic Encryption libraries. Homomorphic encryption makes it possible to compute with encrypted data without decrypting it first; it’s an important step forward in securing user data and other sensitive data.
Social Media
- Social media continues to interact with US politics in surprising ways. K-Pop Stans claim to have reduced attendance at Trump’s Tulsa rally, and Asian-Americans use Slack to help explain racism to their communities.
- Facebook Workplace (their competitor to Slack) allows managers to suppress words like “unionize.” (That feature may have been eliminated at least temporarily, though I find their language vague.)
Cloud
- AWS Snowcone is a device designed for collecting data on the edge of the network. It’s designed to fit into a backpack, so it can be carried by first responders, for example. Like everything else in AWS, it’s a service: you pay per day on-site, plus data transfer.
Miscellaneous
- Apps are on the campaign trail: There’s a whole lot of data collection going on, particularly by the Trump app (though the Biden app is also problematic). “Compare the permissions requested in the Google Play Store.”
- A collaboration between Intel and Berkeley leads to robots that learn to do surgical sutures.
- Google has opened the Maps API to game developers, enabling others to write games that are played across the real world.
