Trying to come up with a theme for the imagery in this post -- too many fake things to report, and we need some visual release in between.
Why not some AI-generated art? It's in the news as of this week because of the first public release of an AI generator that allowed someone to win a local art contest without painting a damn thing.
This is kind of a big deal because for reference, the text-generating tool GPT-3 was deemed too dangerous to release to the public for fear it would, I don't know, topple susceptible governments (maybe we could call them immunocompromised governments instead, to take the edge off).
This newest generation of image generation software is equally as effective, and it's use has been limited to select users, usually researchers and engineers working for either governments (that we don't hear about) or for commercial enterprise (that we do hear about). But now it's open to everyone.
The above image, well, I'll admit I'm not sure how the heck it was made. The artist (the human part at least) titles the work "Stress", but it's likely they used a much more extensive and nuanced text prompt to get the image, and they ran that prompt through a dataset of 5 billion captioned-images and this is what came out. It looks like a much more interesting heart, if not accurate, and perhaps illustrated by MARS1 for a cyberpunk medical school anatomy textbook. ,
Here's a quick timeline of how we got here:
- Google DeepDream - Inception (Jul 2015)
- Nvidia - StyleGAN (2018)
- AI art Edmond De Belamy sells at auction for $432,500 (2018)
- Open AI - DALL-E (2021)
- Discord Server - MidJourney (Mar 2022)
- Open AI - DALL-E 2 (Apr 2022)
- Google - Imagen (May 2022)
- Meta - Make-A-Scene (Jul 2022)
- Stability AI - Stable Diffusion (Aug 2022)
And here's the actual post, about fake things in general, interspersed with AI-generated imagery:
Fake Publishers:
Man arrested in thefts of unpublished books
Jan 2022, Associated Press
Authorities say they’ve solved a publishing industry whodunit with the arrest of a man accused of numerous literary heists in recent years. An Italian citizen working in publishing in London is accused of impersonating others in the industry, using phishing schemes to amass a veritable library of unpublished works. Filippo Bernardini was arrested Wednesday after arriving at John F. Kennedy International Airport. He faces charges including wire fraud and aggravated identity theft. He’s expected to appear in federal court on Thursday. No information on an attorney for him was available.
Fake Whiskey:
With a whiff, 'e-nose' can sense fine whiskey
Apr 2022, phys.org
The real story is that the fake whiskey game is so strong, we need an e-nose to fight it.
And let's not forget the famous wine forger Rudy Kurniawan, he was a wine savant who reproduced rare and exorbitantly priced wines out of a mixture of cheaper wines. What's more mind-mashing -- that he can figure the exact recipe, or that it's even possible to take 5 good wines and mix them in just the right proportion that they taste just like a world-class winner.
[Rudy Kurniawan]
https://en.wikipedia.org/wiki/Rudy_Kurniawan
via University of Technology Sydney: Wentian Zhang et al, The Use of Electronic Nose for the Classification of Blended and Single Malt Scotch Whisky, IEEE Sensors Journal (2022). DOI: 10.1109/JSEN.2022.3147185
Fake Headline:
Spanish police seize huge haul of illegal stuffed animals
Apr 2022, BBC News
Spanish police have seized what is thought to be one of the largest hauls of animal taxidermy in Europe, worth €29m (£24.2m).
Fake Science:
Computer scientists suggest research integrity could be at risk due to AI generated imagery
May 2022, phys.org
It's never the robots, always the humans who are the real problem:
A small team of researchers at Xiamen University has expressed alarm at the ease with which bad actors can now generate fake AI imagery for use in research projects. They have published an opinion piece outlining their concerns in the journal Patterns. (Using a generative adversarial network, obviously)
via Xiamen University: Liansheng Wang et al, Deepfakes: A new threat to image fabrication in scientific publications? Patterns (2022). DOI: 10.1016/j.patter.2022.100509
Real Problems with Fake Science:
Flawed research not retracted fast enough to prevent spread of misinformation, study finds
Jun 2022, phys.org
Makes sense - Retracted papers receive more attention than comparable nonretracted papers even before retraction. One possible reason is that retracted papers may contain sensational or novel findings, which could lead to increased attention, they say.
Using Retraction Watch and Altmetric data databases, the team compared the online footprints of 2,830 retracted papers to 13,599 nonretracted control papers that had similar publication venues, dates, numbers of authors and authors' citation counts. The researchers compared the amount of attention between the two groups of papers measured in mentions in social media and news media databases, and both six months after publication and again post-retraction.
Solutions - Through their hand-labeling process of critical and non-critical mentions, the researchers also found that discourse about retracted papers tended to be more critical overall on Twitter.
via University of Michigan: Hao Peng et al, Dynamics of cross-platform attention to retracted papers, Proceedings of the National Academy of Sciences (2022). DOI: 10.1073/pnas.2119086119
Fake Families:
Used by corporate real estate purchases of residential housing stock, to trick homeowners into selling to a "family" and not a corporation? (This sounds entirely plausible, but I can't find any more info on this beyond an internet forum chat, and aside from the practice by a Dallas-based real estate firm called Centex back in 2006 where they put fake families in the house while they show it, to produce a more "natural vibe" to the place?)
Fake Candy:
Fake 'potentially dangerous' chocolate seized in Oxford Street American candy shop raids
Jun 2022, ITV News London
The council said the raid was part of its investigation into 30 American-style candy and souvenir shops on Oxford Street for business rates evasion amounting to £7.9 million.
The council said it believed that very few of the shops were serving sufficient customers to be commercially viable and were instead being used to avoid business rate bills and possibly commit other offences.
Westminster City Council recently wrote to 28 freeholders urging them to consider the impact of US sweet shops on Oxford Street and has so far seized around £574,000 worth of counterfeit and illegal goods from American candy and souvenir stores.
[Randon Redditor explanation] Covid - only essential businesses could be open, so all the drug fronts switched to become candy stores that were technically food stores that could be classified as essential and continue laundering money. They're american candy because they price things at like 15 pounds for a bag of twizzlers because of the novelty. They're all owned by people and entities from Afghanistan as well, which seems to insinuate they're fronts for heroin. But like the high street in the documentary was like 30% american candy stores with absolutely no one going in and buying anything from any of them. It's pretty wild.
Fake Data:
'Fake' data helps robots learn the ropes faster
Jul 2022, phys.org
How to be a human. We're getting there:
For the rope-looping simulation and experiment, Mitrano and Berenson expanded the data set by extrapolating the position of the rope to other locations in a virtual version of a physical space — so long as the rope would behave the same way as it had in the initial instance. Using only the initial training data, the simulated robot hooked the rope around the engine block 48% of the time. After training on the augmented data set, the robot succeeded 70% of the time.
via University of Michigan: Data Augmentation for Manipulation, arXiv:2205.02886v3 [cs.RO]
https://doi.org/10.48550/arXiv.2205.02886
Fake Lodgings:
Booking.com scam - Tourists descend on north London private home
Aug 2022, BBC News
A woman in north London says she feels victimised after dozens of tourists turned up at her home when her private address was placed on the Booking.com accommodation website.
Travellers from Hong Kong, Saudi Arabia and Los Angeles turned up at the home of Gillian, whose full name is not being given, throughout July.
What kind of fake even is this?
Viagra and Cialis are in certain honey products
Aug 2022, NJ Advanced Media
It's funny because you can bet most of this "honey" is rice syrup:
Honey is one of the most faked foods in the world, and the US government isn't doing much to fix it
And would you look at that, the original announcement from the FDA calls them "honey-based products"
Tainted Honey-based Products with Hidden Active Drug Ingredients
Fake Accents:
Start-up denies using tech to turn call centre accents 'white'
Aug 2022, BBC News
A Silicon Valley start-up has developed technology that can change the accents of call centre workers in real time. The company, Sanas, has told the BBC that its technology could overcome accent-based prejudice and reduce racist abuse faced by staff.
A former call centre agent himself, Mr Narayana said in his experience agents would be abused or discriminated against because of how they sound - abuse the company believes its technology can prevent.
Sanas - Breaking Barriers One Conversation At A Time
Fake Healthcare:
Cutting down vitamin D tests could help lower carbon cost of health care
Aug 2022, phys.org
"Low-value health care", i.e., unnecessary vitamin D tests, because:
"There are high rates of vitamin D testing of healthy individuals who are not at risk of vitamin D deficiency—with previous research estimating that 76.5% of vitamin D tests did not meet any of the Medicare indications for testing. We need to reduce low-value care such as over-testing and overtreatment."
via University of Sydney: Matilde Breth-Petersen et al, Health, financial and environmental impacts of unnecessary vitamin D testing: a triple bottom line assessment adapted for healthcare, BMJ Open (2022). DOI: 10.1136/bmjopen-2021-056997
Fake News (About Spiders)
What fake news about spiders can teach us about the global spread of (mis)information
Aug 2022, phys.org
A similar exercise would allow comparing if the levels of misinformation and sensationalism are the same across a broad spectrum of taxa, testing the prediction if a negative framing by the traditional and social media translates to a lower chance of being prioritized for conservation, and vice versa."
The "Fake Index" perhaps? "The quality of spider information in the global press is rather poor—errors and sensationalism are rampant."
I grew up thinking "eating spiders in your sleep" was an early viral meme created on purpose to measure the network dynamics of the young internet while watching the meme spread. But that's not true. The common answer to where this came from (that you will even find on Britannica unfortunately) is that it was in an article about our susceptibility to misinformation, given as a stupid example of an idea that is easily spread. But that's not true either. Turns out that was a meta-meme, created by Snopes back in 2001, as they pretended to provide the origin story. It's just a perennial urban legend.
via National Research Council of Italy and University of Helsinki's Finnish Museum of Natural History: Stefano Mammola, The global spread of misinformation on spiders, Current Biology (2022). DOI: 10.1016/j.cub.2022.07.026
Fake Influencers:
Real money, fake musicians - A million-dollar Instagram verification scheme
Aug 2022, ProPublica
Fake market always reinventing itself. It's easier to get yourself "verified" as a musician than as an influencer, so they make you a fake album, some fake likes, etc., and you're in business.
Fake Attempts At Correction:
Economists have a method for reducing fake news on social media
Aug 2022, phys.org
"For example, Twitter could limit the breadth of sharing on its site by limiting how many people see any given retweet in their Twitter feeds," he says.
But sharing is a metric for engagement, and engagement is the primary metric for their revenue.
Both Facebook and WhatsApp, two apps owned by parent company Meta that allow users to message each other, have used methods similar to the researchers' model to limit the spread of misinformation.
And that's why it "didn't work", i.e., they didn't really do it.
"If you limit sharing, you could also be limiting the spread of good information, so you might be throwing the baby out with the bathwater and that doesn't really help you," McAdams warns. "Our analysis explores how to strike that balance."
This is the crux of the matter.
via Duke University: Matthew O. Jackson et al, Learning through the grapevine and the impact of the breadth and depth of social networks, Proceedings of the National Academy of Sciences (2022). DOI: 10.1073/pnas.2205549119
No comments:
Post a Comment