DataDownload: Facebook is on disinformation cleanup duty… a tool to help journalists spot fake photos
DataDownload: Facebook is on disinformation cleanup duty… a tool to help journalists spot fake photos A weekly summary of all things Media, Data, Emerging Tech View this email in your browser
As a technologist, I’ve had my share of launches that had a bug or two. Try as you may, they slip through — QA can only do so much. But the Iowa app meltdown is more than that. We’re already facing fears that electric voting machines could be the target of attacks, and this mess underlined those concerns. Then, lost in this story is the fact that the phone lines set to be used were the target of a troll attack, further underlining the danger. Voting is at the core of our democracy — and as technologists we have a responsibility to do, at least, no harm.
As always — please read, comment, and send along your thoughts about how we can work together. Always excited to hear from you.
The NYC Media Lab Must-Read Facebook Will Now Take Down Posts That Spread Coronavirus Misinformation
Facebook is on cleanup duty, scrubbing posts that contain disinformation on the coronavirus outbreak. Usually, posts flagged for propagating false information are simply “downgraded in news feeds,” according to MIT Technology Review. This time the platform is taking a more proactive approach.
Twitter also has its hands full: there have been 15M tweets about the virus in the last four weeks, though there hasn’t been evidence of coordinated attempts to spread disinformation.
1 min read
Read More Tool to Help Journalists Spot Doctored Images Is Unveiled by Jigsaw Google’s Jigsaw unit develops tools to fight censorship, disinformation, and other social issues. Recently they released a free tool called Assembler to help spot doctored photos, including those made by AI. Assembler, which isn’t being released to the public, was trained to detect traces of manipulation:
“The models can detect things like color pattern anomalies, areas of an image that have been copied and pasted several times over, and whether more than one camera model was used to create an image”
4 min read Read More For the Media Google’s Area 120 Launches Tangi, a Short-Form Video App Focused on Creativity and DIY
Google’s in-house incubator Area 120 launched Tangi last week, a TikTok-like social platform for DIY videos. Tangi’s founder Coco Mao was inspired by a visit to her parents, who were watching lots of how-to videos on their phones.
The platform has a feature called “Try It” to let people post their re-creation videos. Apparently, this guac video has the most re-creations:
3 min read
Read More Trolls Conspired to Clog the Democrats’ Iowa Caucus Hotline Amidst the chaos caused by the app used in Monday’s Iowa caucus, a more… traditional network failure went down: 4chan trolls flooded the phone lines meant for reporting precinct tallies. While the calls didn’t compromise the integrity of the information, the demonstration shows how vulnerable our democratic process has become to manipulation — and how unprepared it is to handle it.
2 min read Read More Musicians Are Using AI to Create Otherwise Impossible New Songs
A number of artists have worked with AI to transform their music: YACHT, Arca, Holly Herndon, Toro y Moi, and director at Spotify’s R&D lab François Pachet. While the technology is far from making hits — even Endel CEO Oleg Stavitsky says “AI music is simply not good enough to create a song that you will listen to and be like, ‘I would rather listen to this than Drake’” — it has found a profitable home in the stock music sector.
There is an overwhelming demand for music for podcasts, commercials, and stock videos. Companies like Amper help fill that gap with generated tunes. Endel is similarly serving the background soundscape niche.
11 min read
Read More What We’re Watching Someone Used Neural Networks to Upscale an 1895 Film to 4K 60 FPS, and the Result Is Really Quite Astounding
A YouTuber took the 1896 short “Arrival of a Train at La Ciotat” and upscaled the century-old video to 4k and 60fps. As one of the comments mentions, we’re surprised at the quality of the original!
1 min watch
Watch Now Jobs & Events Event: What is AutoML & Where to Best Apply It
Date: February 17, 6PM-9PM
AutoML is a powerful and trending innovation in machine learning optimization. Expert speakers from AWS & Dataiku explain different areas it can be applied and what use cases are good and poor fits. Register Here.
Event: UX Design for AR/VR
Dates: March 6–7
How do we design for the future of computing? How can we look at past and current trends to design new interactions and interfaces for spatial experiences across virtual and augmented reality? Led by Sam Brewton (Wayfair AR Design Manager, former R/GA) this two-day intensive provides a user-centered design approach for creating spatial computing experiences. The interactive workshops will teach you how to map, design, prototype, and deploy experiences across the mixed reality spectrum. Register Here.
Incubator: Veterans Future Lab Apex program
Deadline: March 10
The Veterans Future Lab (VFL) is accepting applications to its Apex program — a one-year, no-cost, equity-free incubator for early-stage businesses led by those with military ties past and present. Along with a host of resources, Apex offers free housing to six accepted founders relocating to NYC from outside the TriState area. Apply Here. A Deeper Look Comedy Written for the Machines
“The forced acting and desperate re-creation of previously successful formulas make Angelmamii7’s videos into a profound modern drama of people flailing against absurdity.”
A deconstruction of Angelmamii7’s TikTok videos that itself pokes fun at the media’s infatuation with analyzing her content. Author Dan Brooks tried to make sense of Mamii’s popularity, but after hours of absurd content came to the conclusion that performers really don’t why their work gains massive popularity on the platform.
At the root of this confusion is TikTok’s algorithm, a black box with hidden facets mysterious even to the creators: the system serves us what we like, even though we don’t understand why we like it, and triggers its own reward function. (This piece is part of NY Time’s excellent Screenland section — definitely worth a look.)
6 min read
Read More What to Do When You Don’t Trust Your Data Anymore — Laskowski Lab at UC Davis The foundation of data science is built on (what else but) data. If the integrity of that data falls apart after publication, what is a researcher to do? Kate Laskowski did the right thing after finding out her published papers contained faulty data and retracted them.
She details her agonizing experience and the lessons learned during the “crash-course in intensive data forensics.” For a (less-intensive) account of data ethics self-reflection, see PhD candidate Nicholas T. Young’s I Know Some Algorithms Are Biased — because I Created One.
16 min read Read More Transactions & Announcements AI Startup Cresta Launches From Stealth With Millions From Greylock and A16z
Labelbox Raises $25M to Grow Its Data-Labeling Platform for AI Model Training
This email was sent to <<Email Address>>
why did I get this? unsubscribe from this list update subscription preferences
NYC Media Lab · 370 Jay Street, 3rd floor · Brooklyn, New York 11201 · USA