DataDownload: A ring of AI-generated “experts” are publishing propaganda
DataDownload: A ring of AI-generated “experts” are publishing propaganda A weekly summary of all things Media, Data, Emerging Tech View this email in your browser
I was in Central Park today, and I did what all journalists and writers do — I eavesdropped. Little snippets of conversations as people walked by talking, to each other and on their phones. What did I hear? People talking about their lives, travels, schools. Life — but unlike social media nothing was fake. Real-life.
But on the internet, fake is exploding. The Verge has a great long read here. CJR editor-in-chief Kyle Pope says “good riddance to the political coverage that was.”
And MACHINES+MEDIA episode #2 State of the Art in Machine Driven Media features work from Ziv Schneider, Lance Weiler, Anju Kambadur, and Kyunghyun. It’s a great journey of an hour, and I’m proud to moderate these smart folks. Tune in if you can.
Finally — if you’re looking for a podcast — we’re listening to our friend Baratunde Thurston’s “We’re Having a Moment.” He’s funny, in a serious sort of way.
Stay safe, have a fun zoom call, wear a mask.
Reach out with any comments, thoughts, ideas or thoughts. We’re here to be part of the solution.
The NYC Media Lab
Steve@NYCMediaLab.org Must-Read An Online Propaganda Campaign Used AI-Generated Headshots to Create Fake Journalists
The consequence of AI-based attacks were quietly demarcated to deepfake-based reddit posts and academic doomsday scenarios for a number of years… until relatively recently. In 2019, criminals used AI to mimic a CEO’s voice to initiate a fraudulent transfer of $243k. A few months before that, the AP uncovered an “army of phantom profiles” with generated photos on LinkedIn — likely a network of spies aiming to make connections with targets on the platform.
Now, The Daily Beast has revealed a ring of AI-generated profiles that posed as experts, posting large numbers of op-eds “favorable to certain Gulf states” on a range of conservative sites. The Beast uncovered a number of flaws in the profiles — some were mirror images of real people while others appeared AI-generated. The op-eds “argue for more sanctions against Iran, praise certain Gulf states like the United Arab Emirates, and criticize Qatar.”
2 min read
“It’s a lot easier to riff for ten minutes about Trump the climate denier… than it is to understand decades of newsroom failure and corporate influence in the climate crisis.”
CJR editor-in-chief Kyle Pope calls “good riddance to the political coverage that was,” now that the media has shifted its singular focus on Trump and onto subjects that were ignored around the time of the election (in 2015, “twelve hundred people were killed at the hands of police” and Legionnaires’ disease was spreading in NYC — this quickly lost public interest as 2016 rolled around).
The pandemic laid bare the country’s inefficiencies and inequalities, but “as important as the stories have been, every single one of those dynamics was in place and in force before the pandemic, and should have been taken more seriously.” Popes urges journalists to “write and report not just on what the candidates are saying, but on what they should be talking about.”
8 min read
Read More Tech+Media Anatomy of a Fake News Headline It’s frighting how easy it is for an extremist conspiracy theory to weasel its way into the mainstream. Like how elderly Oregonians were algorithmically pushed a Facebook ad about a politician who “Wants Martial Law To Control The Obama-Soros Antifa Supersoldiers”.
The article, originating from political gossip blog Wonkette, wasn’t meant as straight news, but ended up being sucked into the “maelstrom of misinformation” that’s already fueling Americans’ paranoia. The Markup traces the unfortunate, automated path of the headline from blog to news feeds.
6 min read Read More Free Public Data: The Lifeblood of Press Freedom Poynter’s Al Tompkins offers a collection of resources and tips on finding public data. Here’s just a slice — definitely bookmark the full piece as there’s a wealth of useful links:
- “Start wide and large with Google, but learn how to use it in a much more sophisticated way than entering a word in the search bar. Learn how to use ‘allin’ search phrases as well as “related” and “cache” searches.”
- “Then there is Data.gov, the mother of government data sites. It is also a portal to local data sites. “
- “You can hire the National Institute of Computer Assisted Reporting to clean and analyze datasets.”
- “OpenSecrets.org is a place you can go to make sense of the FEC documents. This site makes FEC information easier to search and sort, including campaign disclosures, contributions and spending.”
“Privacy is a civil right. But corporations force you to sign it away every day.”
Sherrod Brown, Ohio’s senior US senator and ranking member of the US Senate Committee on Banking, Housing, and Urban Affairs, guest posts on Wired on his new bill “that takes the burden off of consumers and puts it where it should be: on Big Tech.” In Brown’s own words:
“My bill would drastically scale back the permitted uses of your personal data, banning companies from collecting any data that isn’t strictly necessary to provide you with the service you asked for. For example, signing up for a credit card online won’t give the bank the right to use your data for anything else — not marketing, and certainly not to use that data to sign you up for five more accounts you didn’t ask for (we’re looking at you, Wells Fargo).”
5 min read
Read More What We’re Watching Future of Food: World’s First 3D Printed Plant-Based Steak
Israeli meat alternative startup Redefine Meat is planning on launching a 3D printer that produces plant-based steaks, printing 13 pounds of alt-meat per hour. See the process in action:
2 min watch
IHeartRadio’s We’re Having a Moment is a limited-run podcast hosted by Baratunde Thurston that explores and explains the BLM movement: “When Mitt Romney joins a Black Lives Matter march, when NASCAR bans the Confederate flag, when major cities actually contemplate defunding the police, and when Adidas retweets Nike in support of Black lives, something big is going on.”
Listen Now Virtual Events Virtual Event: Building the Future with 5G
Date: July 16, 2PM-3PM
Chief Product Officer at Newlab, Satish Rao, will moderate a discussion about the possibilities of 5G and how true industry collaboration enables sector-wide transformations. Register Here. A Deeper Look MIT Takes Down Popular AI Dataset Due to Racist, Misogynistic Content
MIT researchers permanently took down the 80 Million Tiny Images dataset, citing the use of “racist, misogynistic, and other offensive” labels. Creators Antonio Torralba, Rob Fergus, and Bill Freeman apologized and urged other researchers to delete any copies they had. The Register originally pointed MIT to a paper — Large Image Datasets: A Pyrrhic Win for Computervision? — which investigated the pitfalls and consequences of large image datasets. The issue stems from the way the dataset was constructed, says Gizmodo:
“80 Million Tiny Images contains 79,302,017 images scraped from the internet in 2006 based on queries from WordNet, another database of English words used in computational linguistics and natural language processing. According to the creators, they directly copied over 53,000 nouns from WordNet, and then automatically downloaded images that corresponded to those nouns from various search engines. Except WordNet contains derogatory terms, and so you end up with results that inadvertently confirm and reinforce stereotypes and harmful biases.”
3 min read
Read More Transactions & Announcements MonkeyLearn Raises $2.2M to Build Out Its No-Code AI Text Analysis Service