‘Biden’ vs. ‘Trump’ and the future of debate

Over on the mega-popular video game streaming and social platform Twitch, the 2024 debate between President Joe Biden and former President Donald Trump has already started.

Well… sort of. If you click here — fair warning, the audio is extremely not safe for work — you will be treated to an AI-powered simulacrum of a debate between the current president and his main rival, complete with uncanny video and voice synthesis.

The joke is that what they’re actually synthesizing is complete nonsense. Running 24/7, the stream’s virtual Trump and Biden riff with a “South Park”-style vulgarity, trash-talking not just each other but the participants in the stream’s chat — the latter being one of the whole project’s most impressive features, as the two deepfaked pols respond in real-time to their viewers.

It’s funny, yes. But it’s also potentially revealing, of a future where we think of politicians first and foremost as the sum of their cultural signifiers in a never-ending competition for memetic superiority. A brief excerpt, chosen for its relative tameness:

TRUMP: “Keep typing away with those tiny fingers while I continue making America great again and kicking Biden’s [redacted] all over this [redacted] debate floor… and as for your [redacted] suggestion of Baby Yoda or the Mandalorian as Biden’s running mate? Pathetic. You know who has real leadership skills? The goddamn AI Trump does.”

BIDEN: “Listen up you bunch of [redacted]. The only thing that compares to Trump’s wall is the wall of [redacted] that comes out of his mouth every time he opens it… The only thing that’s going to pop up faster than Trump’s combover when I become president is my [redacted] for America. We can [redacted] do this, ladies and gentlemen.”

And so on and so forth.

I got in touch this afternoon with Reese Leysen, a veteran of the video-game streaming world and spokesperson for Gaming for Good, the consortium behind the stream whose mission he says is to use monetization to raise money for altruistic social causes like Ebola prevention and children’s safety. What the Germany-based Leysen told me about the project was almost as unexpected as hearing a fake Biden accuse a fake Trump of building a border wall with “Lego bricks and Cheeto dust” — that for all its headline-grabbing vulgarity, the stream’s intent is to push for responsible AI.

“We’ve decided to dedicate our resources to AI and eventually automated general intelligence research and development, trying to come up with architecture and AI that you can trust rather than an intelligent, Lovecraftian machine that isn’t really thinking,” Leysen said.

To that end Gaming For Good has monetized the stream with the option to donate, of course, either to “Biden” or “Trump.” It’s raised nearly $25,000 to date. (Predictably, given the stream’s audience of gleefully transgressive gamers, Trump is outpacing Biden by roughly $3,000.) Leysen says the income will go toward GFG’s own AI development efforts, as they use a combination of PlayHT, OpenAI’s GPT tools, and develop their own homebrewed alternatives to for-profit AI systems like OpenAI and Google.

One could see the incident as not a deepfake-induced crisis of veracity — everything on the stream is clearly, prominently labeled as “AI” and “PARODY” — but the latest in a string of media developments dating back to the Nixon-Kennedy debate. For decades, media critics have seen those debates as a turning point where the medium’s inherent qualities made politicians’ skill as lawmakers less important than their mass celebrity appeal. With the advent of powerful AI tools, is the presence of the actual person relevant at all?

“There are a lot of people for whom politics is really just a tribal war, mostly divorced from policy,” said Jeremiah Johnson, co-founder of the Center For New Liberalism and one of the originators of the pro-Biden “Dark Brandon” meme. “Of course it’s funny to see Trump and Biden talk in weird internet meme stuff, but this stream also seems designed to play with the idea of ‘politics as fandom’… It hits that cheap dopamine rush of dunking on the other side, and Twitch is a platform that’s already very in tune with ‘debate me, bro’ culture and tribal political wars.”

Which is kind of the entire point, from Leysen’s perspective: He told me that the stream is meant to be aggressively apolitical, skirting Twitch’s intense moderation of potentially offensive or controversial content. (He even said his team has regularly seen Twitch staffers watching the stream.) That keeps the focus squarely on the impressive cocktail of AI technology that allows not-Trump and not-Biden to hurl various insults at each other that aren’t fit for printing in a family-friendly newsletter — so impressive, in fact, that Leysen claims a Republican up for election in the ‘26 U.S. Senate cycle approached his team about borrowing it.

If the two pillars of the stream’s hypnotic weirdness are its technical verisimilitude and the symbiotic, hyper-engaged relationship of its viewers with the AI “candidates,” the latter is something that it’s far easier to imagine infiltrating mainstream politics. Popular streamers already broadcasted the 2020 debates via Twitch; cord-cutting and generational trends in media consumption could someday mean that real presidential candidates act out on Twitch what Leysen and his team have made a crude, hilarious mockery of via their techno-wizardry.

Leysen — a dyed-in-the-wool native of the medium — is skeptical. As he sees it, anarchy tends to rule in online gaming communities at the end of the day no matter how large their reach becomes.

“It’s an audience that will just be trolling,” Leysen said. “Of course, you could heavily police the chat. But then what’s the fun?”

Senate Majority Leader Chuck Schumer launched the newest phase of his push for AI regulation today, proclaiming on Twitter that “Congress must join the AI revolution.”

POLITICO’s Brendan Bordelon reported for Pro subscribers on Schumer’s announcement of a “SAFE Innovation framework” for AI, the acronym standing for “security, accountability, foundations and explain,” as well as a series of “AI Insight Forums” in Congress starting this fall.

“There’s such little legislative history on this issue, so a new process is called for,” Schumer told Brendan, warning if lawmakers take “the typical path — holding congressional hearings with opening statements and each member asking questions five minutes at a time, often on different issues — we simply won’t be able to come up with the right policies.”

Actual legislative details have yet to emerge, but the effort matches an ongoing push from the White House to get ahead of AI: Biden’s chief of staff’s office is meeting multiple times per week to develop an AI strategy.

How did “existential risk” become one of the biggest stories in AI?

A MIT Technology Review article published this week explores how that fear has not only spread through the public like wildfire, but torn the field of AI research partially asunder.

“The Overton window has shifted,” writes Will Douglas Heaven. “What were once extreme views are now mainstream talking points, grabbing not only headlines but the attention of world leaders.” Jenna Burrell, director of research at Data and Society, told him she suspects “The threat of genuine regulatory constraints has pushed people to take a position.”

What Heaven points out is that by taking the position that AI might theoretically kill us all, companies also conveniently build up hype about something for which there is no proof or evidence it exists, i.e., an artificial general intelligence: “If you want people to think what you’re working on is powerful, it’s a good idea to make them fear it,” says François Chollet, an AI researcher at Google.