Category: Information

Deepfakes & The Choice to Be Deceived

Immediately after the 2016 election, many Americans discovered just how much fake news they were sharing via social media.  And for about ten minutes, the term fake news had a specific and literal meaning; it referred to fabricated stories made to look like news, and which serve either as clickbait to generate ad revenue or as mischief to fan the flames of political discord.  But then, the president co-opted the term as a way to dismiss any reportage that does not jibe with his myriad, fact-challenged narratives, and fake news no longer means anything at all. 

Now, the unreal is about to get a lot more real—and more dangerous.  The technology known as “deepfakes” enables fairly unsophisticated users to produce video evidence of events that never happened.  As highlighted in this CNN report on the subject, Senator Marco Rubio (R-FL) raises the very plausible fear that, in this next election cycle, we are going see video clips showing elected officials and candidates doing and saying things that are entirely fake, but which look absolutely real. “I believe this is the next wave of attacks against America and western democracies,” Rubio stated in a hearing with the Director of National Intelligence.

And that’s not necessarily the worst effect of deep fakes, at least with regard to news and politics.  As, Hany Farid, a digital forensics expert interviewed in that CNN report, observes, an equal—if not worse—hazard confronts us when people inevitably cry “deepfake” on visual evidence that is indeed factual.  Think about how often President Trump changes his story on just about everything and is then checked against his own prior statements captured on video.  All he, or his spokes-minions, have to do is recite the incantation “deepfake,” and the record is expunged in the minds of millions.  Not that this same folly will not occur among other segments of the electorate, but Trump provides the most obvious, stark, and timely reference in this regard.

Naturally, the anticipation that deep fake technology will be used as a weapon of information warfare leads to the assumption that the remedies will also be technological.  The Pentagon has already called the potential abuse of deepfakes a threat to national security, and Farid makes the logical prediction that social media platforms like Facebook and YouTube will need to deploy deepfake detection software to warn viewers.  But it also stands to reason that faking software will only improve, quite possibly to the extent that it cannot be detected by counter-fake technology.  And even then, can any kind of technical metering overwhelm the psychological instinct to believe what we want to believe?

The truth about our fallibility, as filmmaker Errol Morris’s tells us, is that believing is seeing, and not the other way around.  While images can inform, they just as often lie like crazy, not only because we are hardwired to see what we want to see in recorded images but also because, as Susan Sontag writes,  “…the camera’s rendering of reality must always hide more than it discloses.”

Consider the recent story that began with a viral video clip that appeared to show MAGA hat-wearing teenagers openly mocking a Native American at a rally in Washington D.C. Then, a second video capturing the same events revealed a much broader context that at least alters the original narrative about those kids’ behavior, and possibly undermines it altogether.  Either way, it is impossible to imagine how the addition of deepfakes into this already-volatile environment will not make matters worse.  So, what is the solution to this new form of sophisticated, weaponized information?  

No doubt, there is more than one answer to that question, but, as I’ve opined in the past, I think the only hope is a cultural shift in us as information consumers and not a technological fix on the part of the platform owners.  This might mean, as it did for me, abandoning social platforms as a primary source for “curated” information.  But no matter how we choose to filter information, we have to stop pouncing on every photograph and video clip as evidence to support our “deep stories.”  At the same time, professional journalists must stop trying to keep pace with the shrieking frenzy of social media.

For instance, I initially heard about that D.C. clash on CNN, when they cited the first viral video as evidence that a mob of teenagers had indeed assaulted a Native American elder.  The anchor reporting the story even editorialized with a scornful word or two about the kids’ conduct.  But then, CNN followed up, reporting that a second video shows a “different side of the encounter,” and they hosted an interview with Nathan Phillips (the Native American), which also skews the story considerably from the way it was originally reported.  But does CNN’s follow-up do enough to build any kind of consensus around the truth?

When I first started this blog, the trending videos at that time were coming from the cellphones of Occupy Wall Street attendees, usually depicting apparent acts of police brutality against allegedly peaceful protestors.  Clearly, such incidents did occur, but at the same time, the omnipresence of cameras—especially at a movement that quickly devolved to activist tourism—helped to foster an illusion that the people’s images are the “real” truth, even to the extent that citizen journalism has eroded trust in professional journalism.  

This is not to say that amateur video cannot tell us anything.  Surely it can.  But the inexorable deployment of deepfakes, which will probably be most effective when disguised as citizen journalism, will be all the more hazardous if we cannot trust real journalists to provide context, corroboration, or correction for what we think we’re seeing.  In this regard, CNN’s own deepfakes reporting might serve as a cautionary tale to its main news desk (and every other news organization) that the visual “evidence” they obtain via social media and other outside sources should be treated with a level of scrutiny as though it were mere rumor.  And, as consumers, we should begin to do the same.

Photo by kiosea39

On New Models, Journalism, and Digital Advertising

It was encouraging to see our most prominent millennial Member of Congress, Rep. Ocasio-Cortez (D-NY) recognize the link between a healthy democracy a professional class of journalists. On Friday, presumably in response to the startling number of layoffs at BuzzFeed, @AOC tweeted this:

True to form, Mike Masnick of Techdirt replied:

It is ironically quaint at this point to see anyone, even Masnick, still using the “buggy whip” metaphor.  I mean could the term beat a dead horse be any more appropriate?  The buggy whip was always a stupid reference because horse-drawn vehicles are, in fact, obsolete, while the content that big tech companies exploit and devalue (like journalism) is clearly still very useful and in demand.  

Several years ago, the “adapt to new models” narrative was just dumb magical thinking.  But today, we have ample evidence to call this talking point a demonstrably failed proposition.  I guess it’s good that Masnick did not suggest journalists should tour, sell merch, or find new ways to connect with their fans; but still, Mike should go lie down by his dish and think about what he’s done.  

There may be new models in the sense that we enjoy new ways to access and experience content—be it news or entertainment—but there are no truly novel economic models to support the production of content in a free market.  The revenue needed to pay reporters, writers, etc. comes from consumers or it comes from advertisers.  Everything else is alchemy.  And while there are certainly many other factors external to Facebook and Google that have changed the nature of journalism and our relationship to it, the market reality for news and other content creators is that the major internet companies systematically poisoned both revenue streams.

First, the industry laid siege to the principles of copyright and promoted a faux-populist (frankly childish) message that all content must be free.  Then, they helped fulfill the promise of free by erecting giant tollbooths that siphoned off the lion’s share of the available ad revenue, which would otherwise go directly to content creators like journalists.  It’s funny that the free-content, anti-copyright crowd tend to mock as anachronistic any news organization that would presume to put up a paywall, but that’s exactly what Facebook is—a paywall.  No, we don’t pay to use it, but the content creators pay with the lost revenue they rightly earned.

It is especially funny (or sad) that Masnick would bring out a variation on the adapt message in context to BuzzFeed, which IS a new model.  It was built as an online-only platform that would be free to consumers, and it was designed with social media in mind.  Yet, as the New York Times reports, founder Jonah Peretti believes the solution to the Facebook/Google problem may be a merger of several digital news networks into a group that can negotiate better terms for ad-revenue sharing.

But, again, notice how there’s no “new model” there.  It’s just an old model called advertising now dominated by two massive companies.  And the fact is that news media companies have adapted, although in the ever-changing landscape of platforms like Facebook, it is probably more accurate to say that they have reacted in ways that are of little value—economic or social—to the purpose of journalism.

In October of 2018, Alexis C. Madrigal and Robinson Meyer, writing for The Atlantic, reported that several news companies laid off dozens of reporters, mostly writers, to make room for video production resources in an effort to capitalize on Facebook’s new video initiative.  Citing a lawsuit pursuant to Facebook’s allegedly misrepresenting the data on video impressions for advertisers, the authors write…

During the period of purported wrongdoing, from July 2015 to June 2016, journalists and newsroom leaders across the country worked to cover an unprecedented presidential campaign in an information landscape that Facebook was constantly, and erratically, transforming. Even if, as Facebook argues, it did not knowingly inflate metrics, it set up new and fast-changing incentives for video that altered the online ad market as a whole. 

So, even if adapting to video had proven remunerative for news companies, this is still not a good environment for journalists, or for the public that relies on their work.  News organizations should focus on doing the best job of reporting the news, not figuring out how to navigate the opaque and erratic landscape of Facebook.  As I say, that’s not adapting, it’s reacting; and that same Atlantic article cites one example that makes this point.

There is something seriously flawed in the narrative that BuzzFeed potentially broke an important story this month about Michael Cohen’s testimony and then had to decimate its national news team last week–but that, in 2016, they spent resources making a viral video featuring two employees exploding a watermelon.  That is adapting to new models? Hard-news supported by an old Gallagher joke?  And it didn’t even work.  “BuzzFeed never repeated its success,” write Madrigal and Meyer. “But that didn’t stop reporters from being taken off the line of duty, while a promotional video of water being poured on permeable concrete racked up 100 million views.”

Meanwhile, as intermediaries collect the ad revenue that content creators like journalists generate, the advertisers themselves may be getting a raw deal themselves.  Facebook’s allegedly fraudulent reporting of video-view metrics is consistent with other evidence suggesting that trouble in the digital advertising market may be far from over.  As cited in a recent post, Max Read of New York Magazine tells us that a staggering amount of the internet, at any given moment, may be fake.  Read writes …

Studies generally suggest that, year after year, less than 60 percent of web traffic is human; some years, according to some researchers, a healthy majority of it is bot. For a period of time in 2013, the Times reported this year, a full half of YouTube traffic was “bots masquerading as people,” a portion so high that employees feared an inflection point after which YouTube’s systems for detecting fraudulent traffic would begin to regard bot traffic as real and human traffic as fake.

What all that means for advertisers, of course, is that they’re not getting the impressions they’re paying for, let alone the quality impressions digital ad sellers continue to promote. If this is the case, it implies that another reckoning may be at hand between the major advertisers and Facebook and Google.  Wouldn’t it be interesting if the solution for both advertisers and news organizations is that the brands return to buying more media from the news sites themselves rather than the intermediaries?  Yeah, I know.  It’s an old model.  But it worked pretty damn well.

Robot image by frescomovie

Mach (digital) Tuck?

Not long after I wrote a post suggesting there is little difference between naive human engagement and bot engagement on policy issues, a couple of things happened.  One was the publication of a story by Max Read in New York Magazine reporting that a substantial (though hardly surprising) amount of material and people on the internet are fake.  The other thing was a recent get-together with my best friend of 30 years, who told me that his eighth-grader son said he was worried that “Article 13 is going to destroy YouTube.”  Perhaps most tellingly, the kid complained to his father, “The [legislative] language is too vague!”

Now, I love my friend’s kids as I love my own, and I hope they and their contemporaries become engaged citizens as they grow up; but none of them is yet qualified to have an informed opinion about the precision, or lack thereof, in the language of bills proposed at the European Parliament or any other legislative body.  The broad irony here is hard to miss:  In contemporary, digitally-distracted America, we can only hope the next generation learns about the constitutional separation of powers and other rudimentary civics, while YouTube videos frighten them about “hazards” lurking in arcane proposals they cannot possibly have the background to understand.

This is no way to run a liberal democracy, yet my friend’s son in this scenario, to no fault of his own, has been turned into a bot.  Fortunately, his father doesn’t generally take the policy views of his children at face value, but you can bet that more than few parents are apt to be less attentive when it comes to some obscure (let alone European) bit of cyber-policy.  Their kids announce,  “Hey, XYZ will destroy the internet,” and the next day, a meme with the same message just happens to appear on Facebook, Mom or Dad shares it, and civilization is destroyed one guileless click at a time.  

Of course, it isn’t one click at a time.  It’s millions of clicks that, apropos the aforementioned New York article, do not even represent real people in many cases.  In fact, Max Read states, “Studies generally suggest that, year after year, less than 60 percent of web traffic is human; some years, according to some researchers, a healthy majority of it is bot.”  

If that statistic is off by even a wide margin, that is still a hell of a lot of bot, which leads me to the following question:  If we combine total bot + humans acting like bots + hackers & other disinformation brokers, how close are we to approaching maximum inescapable bullshit?  We might also add just plain bad reporting under brand names like this piece on forbes.com about Article 13, which refers to the non-existent “FAIR USE Act” — an error gross enough to disqualify the author from commenting on copyright law at all.  But here we are.

It was the same best friend mentioned above who, when we were college freshmen, taught me the aeronautical term mach tuck.  Put simply, this is a hazardous condition that occurs, usually in subsonic aircraft traveling faster than they’re supposed to, when the airflow over the wing nears, or even exceeds, Mach 1.  This causes the nose of the plane to pitch downward, and in an aircraft not designed to fly at or above the speed of sound, correction may not be possible.  Thus, the downward pitch naturally causes the plane to accelerate, which exacerbates the problem by increasing the windspeed over the top of the wing.  This vicious cycle called mach tuck can plunge the aircraft into an unrecoverable dive.  

Obviously, I took that aerodynamic detour to ask whether the “information age” has achieved—or could soon achieve—mach tuck?  Is the flow of garbage streaming so quickly, and accelerated by forces beyond our control, that we are about to enter—or are already in—an information nose dive from which we cannot recover?   

Optimistically, I don’t think so.  While it is not entirely clear that the pros of digital engagement sufficiently outweigh the cons, the pros are still evident and abundant.  No one can, or should, doubt that there is a wealth of useful, credentialed, instantly-accessible information online.  The challenge, though, is that the major social platforms—which happen to be owned by the wealthiest and most politically-influential corporations—distort information through a mosaic of visual stimuli that probably overpower critical thinking.  Assuming that’s true, things may be about to get worse before they get better.

I suspect 2019 will be the year a lot more people hear about “deepfakes,” a software developed by Google, which enables even a modestly-skilled individual to effectively “skin” any face onto any body in a video clip.  Both “deepfakes” and the more elaborate CGI capabilities of motion picture suppliers have raised new questions about publicity and a celebrities’ rights to control the use of their likenesses in mainstream fare.  And, yes, not at all surprisingly, the likenesses of popular actresses have indeed been grafted into pornographic scenes in which they did not appear.  

But while these issues for mainstream performing artists are debated as a matter of policy and contract law in the coming year, stay tuned for information warfare to get a lot uglier through the use of these technologies.  New darling Democrat Alexandria Ocasio-Cortez was throwing molotov cocktails at nuns?  Of course she was, here’s the video.  Is that Rep. Doug Collins giving a speech at a Nazi rally?  Roll that clip.  And all of it hosted on YouTube which doesn’t have to remove any of it because…free speech?  

The old cliché that “seeing is believing” has always been a duplicitous axiom.  Just about any major critic who has ever written about photography will tell you that seeing may result in believing but that this should not be confused with seeing the truth.  “…the camera’s rendering of reality must always hide more than it discloses,” wrote Susan Sontag in 1973, long before the technological capacity for amateurs to create realistic, moving pictures, depicting real people in scenes for which they were never present.

Of course, if fake video clips like I describe are deployed en masse, there is the possibility that this could trigger a healthy skepticism for believing what we see.  Presumably, this will depend on the degree of subtlety employed by the manipulators, and it is worth noting that the hackers at the St. Petersburg-based Internet Research Agency have been described as both subtle and sophisticated in their use of disinformation on social media. 

Of course, if rampant fake video were to induce new skepticism, this implies a potential  new hazard—that we no longer believe what is, in fact, true (i.e. mach tuck).   These are the kind of challenges that companies like Google should be helping to address instead of spending their vast resources to scare the kids about comparatively modest proposals like Article 13 in the EU Single Market Directive. 

The irony that these companies invoke free speech in their efforts to protect their own revenue could not be more pellucid as their platforms and policies literally help to unravel the very reason speech was a protected civil right in the first place—a hope among a handful of 18th century idealists that the electorate, while always debating what should be, might at least find common ground in what is.