The TikTok Gap Isn't About TikTok

It's older pols trying to steal younger activists' audiences, and whether or not you are "viderate." Plus, more on how generative AI may warp politics, this time with speech cloning.

In December 1963, Bob Dylan was given the Tom Paine Award by the Emergency Civil Liberties Union, a prominent group of liberal New Yorkers, at their annual Bill of Rights Dinner. By then, Dylan had written classic protest songs like “The Times They Are a-Changin,” “Masters of War,” and “Blowin in the Wind,” and he had sung at voter-registration rallies in Mississippi and at the March on Washington. The older heads of the ECLU, which had banded together a decade earlier after the ACLU refused to defend the rights of people accused of subversion under the 1950 McCarran Act, had every reason to expect that their embrace of this voice of a new generation would be reciprocated. Instead, Dylan—who got drunk before he got on stage in front of the more than 1000 guests—mocked his adoring audience.

“You people should be at the beach,” he declared. “You should be out there and you should be swimming and you should be just relaxing in the time you have to relax. (Laughter) It is not an old peoples’ world. It is not an old peoples’ world. It has nothing to do with old people. Old people when their hair grows out, they should go out. (Laughter) And I look down to see the people that are governing me and making my rules – and they haven’t got any hair on their head – I get very uptight about it.” Boos rose, and even hisses after he declared that he “saw some of myself” in Lee Oswald, the man accused of assassinating John F. Kennedy a month earlier. All in all, it was not a successful evening.

This story came to mind as I read the second in a series of reports from the Cooperative Impact Lab titled “Filling the Progressive TikTok Gap.” In it, Ben Resnik, Kate Gage, Valeria Sosa Garnica and Oluwakemi Oso share some valuable lessons that they learned from a four-month push in 2022 to test a variety of messaging strategies on the highly viral platform. They’re doing timely and important work. As they note, “Across the space, from TikTok newcomers to seasoned professionals, virtually no progressive groups feel they have a grip on what effective political TikTok content looks like, how to generate it, or how to have it support organizing strategy consistently. As established digital tactics like Facebook ads get more expensive and less effective, the knowledge gap on how to speak online is becoming downright dangerous.” [Emphasis in the original.]

When I first saw the lab’s new report, which was posted February 23^rd, I have to admit I didn’t pay much attention. It looked too much like their first study, which came out in late November. If there’s a “progressive TikTok gap,” I thought to myself, surely that’s just a tactical problem for the progressive messaging community. They’ll solve it the same way they’ve figured out how to communicate on other platforms, through a lot of trial and error and copying of what works. And that indeed is part of what Resnik and crew offer in their new report: hard data on which kinds of messages and messengers connect best. Offerings of strict information (how to vote, for example) do worse than narrative content (why to vote). And using “direct-to-camera storytelling, skits, memes, and other techniques to make the information contained more relatable and enjoyable” also make narrative content more effective. There’s some additional actionable advice on things like whether to peg content with issue tags, the value of paid promotion, how to navigate TikTok’s political advertising guidelines, and whether resource-constrained organizations should even venture onto TikTok if they can’t investing in making high-quality, high-frequency content from the start, since the app’s algorithms appear to quickly relegate new accounts that perform poorly to permanent cyber-Siberia.

The big takeaway, though, is this: the knowledge gap of how to use TikTok for politics is a symptom of a generation gap in how we connect and communicate. The Cooperative Impact Team seems to get this, as they argue in the conclusion of their report that “TikTok is not just about one platform — it’s about rethinking our entire relationship to persuasion and messaging.” They call for earlier investments in influencer recruitment, because it “requires time to establish long-term relationships with mission-aligned content creators.” And they call for more research on everything from best practices in creator recruitment and content production as well as how to better measure narrative change.

This is all for the good. As they are among the first generation of political practitioners to grow up in data-driven environments, I get why Resnik et al are arguing that we’ve got to study this harder. I suppose if a young Dylan had shown up at, say, Netroots Nation and insulted the poohbahs in the room for their age and squareness, they might also say that we need to rethink our entire relationship to persuasion and messaging—without recognizing the near futility of the endeavor. Using younger people’s popularity to drive older people’s projects is inherently problematic, because the nature of youth is to rebel, carve out autonomous space and try to stand on one’s own. You can only “use” a talented messenger for so long before they bite the hand that wants to give them a fancy award instead of the actual change they’re demanding. “Mission-aligned content creator” is another way of saying independent activist-organizer; maybe they have minds and wills of their own? As the man wrote, “your sons and your daughters are beyond your command.”

Of course, older, more established people in the world of politics have been trying to tap the energy and creativity of younger people, without giving them actual agency, for a very long time. It’s not just old Upper West Side fellow travelers and Dylan, it’s Richard Nixon posing with Elvis Presley, Jimmy Carter palling around with the Allman Brothers, Willie Nelson and yes, Dylan, Bill Clinton adopting Fleetwood Mac’s “Don’t Stop Thinking About Tomorrow” as his theme song, Barack Obama taking online questions from YouTubers, the DNC shlepping a bunch of Instagrammers and TikTokers to shoot selfies with Joe Biden last fall, all the way up to my very own Congressman Jamaal Bowman (D-NY) doing a rap song with newly elected Rep. Maxwell Frost (D-FL), the youngest member of Congress, at this past weekend’s Democratic issue conference.

And while it seems like older people in politics have nearly all the power (money, connections, experience), younger people like the TikTok influencers currently being coveted for their audiences have one thing the older generation mostly lacks: videracy. That is, they are highly literate in communicating using video. My longtime partner at Personal Democracy Forum, Andrew Rasiej, tried to coin this term back in the early days of YouTube, but it never really took off. It may not stick now, either, because of how fast the communications environment keeps changing (more on that topic below). But successful communication via digital platforms today requires the kind of fluency that can only come from immersion in their use. And if you are someone who is really good at connecting with others over TikTok, maybe you’re going to want to use that skill for something more than the incremental-at-best and gridlocked-at-worst change that is currently on offer from the powers that be.

More on AI’s Impact on Politics        

Continuing my recent series of posts on how ChatGPT may transform the political arena, it’s worth taking note of the appreciable increase of cloned speech of public figures appearing in the wild in the last few weeks. See for example this fairly funny and well-produced clip of President Biden supposedly talking about his preferences in weed. That’s not an impersonation; it’s a piece of synthetic audio most likely made by collecting a minute or so of Biden’s actual speaking voice, loading it onto Eleven Labs speech synthesis platform, and then inputting the desired text. (In this case, the maker also added in some realistic audience sounds to make it seem Biden was speaking before a crowd.) Until recently, this was somewhat complex to do; now Eleven Labs will let you use its VoiceLab for just $5 a month and all you need to understand is how to record and upload an audio clip. When using a new technology becomes as easy as paying a small fee and attaching a file, usage is bound to explode.

Sam Gregory, the executive director of Witness, has been working intensively on the problem of synthetic media from before anyone had even heard of the term “deep-fake.” He agrees that there’s been “a clear increase in the use of cloned and manipulative speech” of late. “Certainly ElevenLabs tool sparked a lot of malicious usages,” he told me, “but in addition you're seeing more visibility on TikTok (e.g. the Biden talks to Obama TikToks, e.g PossumGirl) of voice cloning. One of the areas that was most visible in the recent Brazilian elections was voice manipulation and editing (sometimes manual, sometimes automated).”

It seems to me that, when you add in ChatGPT’s new API, which is already enabling big platforms like Snapchat to power conversations with users and Instacart to offer food suggestions, we're very close to (if not already) past the point where someone could record your voice off a podcast or video talk you did, clone it on Eleven Labs, and then call your friends and relatives and not only really sound like you, but seemingly carry on a conversation. This could be used in all kinds of ways in politics, including: a new level of robocall sophistication, where instead of politicians sending pre-recorded calls, their bot calls constituents one by one to ask them how they're doing and find out what they care about; a new level of robocall sophistication, where instead of politicians sending pre-recorded calls, their bot calls constituents one by one, targeting them by pre-inputted preferences based on their social media and consumer data profiles and serving up microtargeted and manipulative pitches for their vote; a new level of robocall defense, where instead of real people, people's answering machines respond to calls from volunteer phonebankers and our ability to reach voters by phone crashes. Not to mention all kinds of potential dirty tricks one could make to make a speaker look bad.

Gregory agrees that the opportunities for personalization of political messaging have been opened up both for reasonable goals, such as “personalized interactions at scale with transparency,” as well as evil, manipulative ones. “We have a massive potential to increase the volume of increasingly personalized deceptive audiovisual communication,” he says. For all we know this won’t hit politics as quickly as other sectors; one can easily imagine a spam factory in the developing world figuring out how to make a completely new generation of programmatic dialers that, instead of spamming your contact list with a fake email that seems to be you asking for emergency help on your overseas trip, instead is a voice mail that actually sounds like you asking for the same help. Not crazy: here’s Motherboard privacy reporter Joseph Cox using an AI-generated simulation of his voice to get through his bank’s security system (say goodbye to voice-printing security systems).

Lawmakers ought to be hitting the brakes to allow some time for all of society to consider the implications of this powerful new technology, or at least making clear that companies like OpenAI will be liable for its misuse.

—Related: Microsoft’s Bing Chat, which is also powered by ChatGPT, has a celebrity mode that allows the bot to impersonate the speaking style of various celebrities, including such as Elon Musk, Tom Cruise, Taylor Swift, Kevin Hart, Dwayne Johnson, Beyoncé, Oprah Winfrey, Tom Hanks, Diddy, Barack Obama, and Lebron James.

—Also related: New services are also offering easy ways to make synthetic videos of AI-generated people saying whatever you want, as Alex Pasternack reports for Fast Company; last year a series of videos appeared on WhatsApp featured a number of fake people with American accents awkwardly voicing support for a military-backed coup in Burkina Faso.

—Further reading: On TechPolicyPress, Matthew Ferraro, counsel at Wilmer Hale, has a solid ten-point list of the legal and business risks already arising out of the use of generative AI.

Odds and Ends

—Here’s a real-time map of the hundreds of protests taking place across Israel as the pro-democracy movement there hits its ninth week of massive participation. An estimated 400,000 people turned out across the country last weekend; that’s five percent of its population—imagine 16 million Americans on the streets week after week.

—Best-selling author and futurist Yuval Noah Harari has added his voice to the Israel democracy movement; here’s a short compilation (with translation) of his remarks from last Saturday’s mass demonstrations.

—Yes, New York Times columnist Thomas Friedman did just call Senate Majority Leader Chuck Schumer (D-NY) a “useful idiot” for hugging beleaguered Israeli Prime Minister Bibi Netanyahu recently:  “I saw pictures yesterday of Chuck Schumer smiling, yucking it up with his old pal Bibi and in the parliament. And my attitude is you are a useful idiot Chuck, you are a useful idiot. Just to get some picture back home and the Jewish Week in New York, you know that might get you a few votes, like would you be yucking it up with Kevin McCarthy right now? Would be up yucking it up with Tucker Carlson. Well, what are you doing yucking it up with him? Useful idiot.”

—Apply: Higher Ground Labs has released its 2023 investment thesis (focus on unmet ecosystem-wide needs) and is looking for applications to its accelerator. Notably NOT on its radar: anything having to do with Web3 or blockchain. Good riddance.

Deep Thoughts

AI critic and computational linguist Emily Bender has some really important questions for all of us, including “’Hey, look, this is technology that really encourages people to interpret it as if there were an agent in there with ideas and thoughts and credibility and stuff like that. …“Wait, why are these companies blurring the distinction between what is human and what’s a language model? Is this what we want?”

End Times

Gigi Sohn has withdrawn from her nomination to the FCC; it’s a victory for the cable and media industry lobby and the politicians in its pocket, and a huge loss for consumers.