🤥 Faked Up #7

YouTube goes Birdwatching | America's collateral damage | There are no AI candidates

Happy Wednesday.

Faked Up #7 is brought to you by tortoise lies and sloppily stolen satire. The newsletter is a ~7-minute read and contains 65 links.

Top Stories

YOUTUBE GOES BIRDWATCHING

I was probably the Google employee who tweeted the most about Birdwatch1, X’s 3-year-old feature that appends crowdsourced corrections to tweets. So I’m not going to lie, I’m a little peeved that YouTube waited for my departure to copy the feature and launch Viewer Notes.

But this is not about me. This is about YouTube users who, according to the company’s mocks, will NO LONGER BE MISLED ABOUT GIANT TORTOISES!

Just kidding, this is obviously part of YouTube's strategy to prepare for the US 2024 election following its partial rollback of election denialism policies.

YouTube’s Notes are remarkably similar to X’s, all the way down to a “bridging algorithm” meant to promote helpful comments. Hopefully, YouTube’s imitation of X will extend to making notes available in a downloadable format.

A lot of work will have to go into making these notes useful. A helpful note, YouTube writes, "cites high-quality sources." YouTube will hope its users define high quality the same way as Google’s guidelines, but that’s a tall order. Just yesterday a preprint dropped analyzing the sources used in Birdwatch. Turns out the single most cited source is…Twitter.

The incentive structure will also be crucial. X's notes have been largely driven by counterpartisans eager to correct their ideological adversaries. Unless YouTube does something dramatically different, I expect this will happen in Viewer Notes too.

Still, it’s a worthwhile experiment that I’ll be following closely.

AMERICA’S COLLATERAL DAMAGE

In an explosive investigation, Reuters claims that at the height of the COVID-19 pandemic, the U.S. military conducted an influence campaign targeting Filipinos with anti-Chinese messaging. The campaign, run by the military’s psychological operations center in Tampa, used inauthentic accounts on Twitter to push the hashtag #ChinaAngVirus (#ChinaIsTheVirus) alongside claims that Chinese-supplied vaccines, masks and test kits were unsafe.2

This report also provides significant additional details on a previously reported operation run by the US military on Facebook targeting Central Asia.

Reuters claims that the psyops were conducted despite opposition from State Department officials in the Trump administration and a prohibition from the the Biden administration.

When asked about the consequences of the anti-Sinovac campaign on the Philippines, which at the time had an extremely low inoculation rate and 24,000 deaths from the virus, a senior military officer had this to say:

“We weren’t looking at this from a public health perspective … We were looking at how we could drag China through the mud.”

50 MORE

Baccus Marsh grammar school is the latest school community to record an incident of AI-generated nudes. According to the ABC News, a teenage boy was arrested following the discovery of AI-generated fake nudes of approximately 50 students attending the Melbourne institute.

According to my tracker, this brings the number of reported underage victims from AI undresser apps in school settings up to 174 across seven countries. This undoubtedly an undercount given that it relies on these incidents being (a) detected and (b) reported by news outlets I track.

THERE ARE NO AI CANDIDATES

Steve Endacott and Victor Miller don’t appear to have a bunch in common. Endacott is a British businessman with a background in the travel sector; Miller is an American facilities technician at the Laramie County Library System in Wyoming. Yet both made headlines globally for running for office claiming they’ll put an AI in charge.

It is true that Steve AI is on the ballot for MP in Brighton Pavilion and VIC for mayor in Cheyenne. But both campaigns are at most AI-augmented gimmicks.

Miller claims to be the AI’s “meat avatar” and goes around with a Bluetooth speaker around his neck to project his chatbot’s answers. But local reporter Jared Gendron confirmed to me that at a recent candidate Q&A “Miller did all the speaking.” VIC was also booted off by OpenAI and its inclusion on the ballot is being investigated by county lawyers following an inquiry by the Wyoming Secretary of State.

The UK has notable precedent with pseudonymous candidates, which can be allowed under a “commonly used name” policy or by registering the name as a party (thanks a ton to Will Moy for helping me decrypt UK voting laws).

So Steve AI may stay on the ballot. But reading Endacott’s campaign website, it seems the AI will aggregate crowdsourced policy ideas that will then be voted on in a garbled attempt at direct democracy.

Steve AI is also clearly a marketing ploy for Neural Voice, the company which built his avatar and for which Endacott is an investor and Chairman. In one LinkedIn post, Endacott invites his #travelandtourism network to “imagine what you could [do] with this tech” and suggests that after his campaign Neural Voice “might get a bigger investment.!!!”

Steve AI may also face technical hurdles. When I asked it on Monday, it claimed to be operating on GPT-3, which may violate OpenAI’s policies. The Neural Voice website also claims its tech leverages “the sophisticated conversational capabilities of ChatGPT” (I emailed OpenAI on Monday for comment but have not heard back3).

A FAMILIAR PLAYBOOK

The rector of the Federal University of Rio de Janeiro (UFRJ) accused Brazil’s parliament of following “a familiar playbook used in other countries by far-right agents aiming to censor science through parliamentary or congressional action.” This is in response to a request for the UFRJ research center NetLab to testify at the Public Security and Organized Crime Committee on its work tracking misinformation and scams related to the Rio Grande do Sul floods.

NetLab has also been accused of bias by Meta, in a legal response to a fine imposed on the company in 2023 for running scam ads about the welfare program Desenrola Brasil. A coalition of NGOs and researchers signed an open letter decrying Meta’s defense as an “attack on scientific research work.”

5 YEARS OF POFMA

The Singapore Straits Times reviewed the 163 directives issued by the city-state’s government under its controversial anti-fake news law since its inception in 2019. The law, known as POFMA, frequently targeted opposition leaders and required corrections appended to their online posts. The most frequent topic of the posts targeted was COVID-19, followed by the Ridout Road rentals controversy.

CALIPHATE NEWS NETWORK

The Institute for Strategic Dialogue reports that groups supporting Islamic State ran fake CNN and Al Jazeera channels on Facebook, X, and YouTube in order to spread their propaganda through “faux-branded newscasts.” The ISD notes that “the pages had small followings,” but “the networked nature of IS supporters online created momentum,” with supporters downloading the videos and r-uploading them across multiple pages. Wired has a good write-up of the report.

ALL’S GOOD AT ALLOD

The New York Times reports that America’s Last Line of Defense — a Facebook page that has specialized in rage-baiting anti-woke types with satirical fakes since at least 2017 — is getting record engagement on Facebook.

Chasing the algorithm, ALLOD has pivoted from explicit politics to culture war topics that involve celebrities. This is a typical post:

In the comments, the page notes that the controversy is made up.

But many do not seem to be in on the joke.

Anyhoo, I’m most impressed that amid all this, Blair is complaining about the effect of generative AI on his oeuvre. A particular bugbear is the SpaceXMania group of pages, whose founder is based in Pakistan, and whom Blair accuses of plagiarism:

“My material, my cast of characters, my keywords, my hot buttons — they take everything […] They put it into an A.I. program, and it just spits out headlines. There’s nothing original about any of it.”

I asked PolitiFact editor Katie Sanders about this whole circus. She told me that stolen satire has been around for a while. Done sloppily, however, it “presents new challenges” by making the border between what is fake and what is satire even more porous. This matters because fakes are penalized by Meta, but satire is not.

Headlines

1. Far-right parties employed generative AI ahead of European Parliament elections (DFRLab)
2. Microsoft Chose Profit Over Security and Left U.S. Government Vulnerable to Russian Hack, Whistleblower Says (ProPublica)
3. Hackers Target AI Users With Malicious Stable Diffusion Tool on GitHub to Protest 'Art Theft' (404 Media)
4. Experts worry X's new policy on AI-generated adult content will lead to more deepfake porn (Yahoo News) with Senate passes bill to curb AI deepfakes (ABC 27)
5. Public attitudes towards the use of AI in journalism (Reuters Institute for the Study of Journalism)
6. Google AI Gemini parrots China’s propaganda (VOA News)
7. AI experimentation is high risk, high reward for low-profile political campaigns (AP News)
8. Effects of Antivaccine Tweets on COVID-19 Vaccinations, Cases, and Deaths (arXiv)
9. Top 10 Generative AI Models Mimic Russian Disinformation Claims A Third of the Time, Citing Moscow-Created Fake Local News Sites as Authoritative Sources (NewsGuard)
10. GOP Candidate Claims Incest Arrangement Audio Is a ‘Deepfake’ Meant to Smear Him (404 Media)
11. A major disinformation research team's future is uncertain after political attacks (NPR)
12. AI Images in Google Search Results Have Opened a Portal to Hell (404 Media)
13. Fake anti-Ukraine celebrity quotes recently surged on social media (The Record)
14. Meta has created a way to watermark AI-generated speech (MIT Technology Review)

Before you go

A competent procurement team is essential to every operation4

“There are at least three Facebook pages devoted to AI-generated images of Keanu Reeves with animals”

I love this.

AI has got nothing on my dad jokes

More BNN chicanery

1 Birdwatch is now known as Community Notes because nothing is forever on the app formerly known as Twitter.

2 The Sinovac vaccine may have been less efficacious than others but was deemed nonetheless safe and useful at reducing the spread of the disease.

3 OpenAI confirmed to CNN in an article published Wednesday that it took down AI Steve’s account.

4 Following publication, I discovered the tweet in the screenshot was gibberish and probably a troll. I regret the error.