There's a culture war brewing over "woke AI" but The Atlantic can't see it

Right-wingers are up in arms over an imaginary threat once again

A strange article went up on The Atlantic’s website yesterday, asking, “Where is the AI culture war?” In it, writer Jacob Stern asserts that

somehow, in an era when kitchen appliances and children’s books can erupt into new fronts in the culture war virtually overnight, AI has so far managed to remain strikingly nonpartisan.

Really? If you can’t find evidence of a culture war brewing over AI, you may be looking in the wrong places. For there is already a movement developing on the right against what is invariably caricatured as “woke AI,” based on the premise that evil liberal overlords have taken control of AI and are using it in a propaganda war against the right. The movement is still nascent, but it is very much real.

While “woke AI” hasn’t yet emerged as a culture war villain for the right in the way that, say, drag queens and the COVID vaccines have, it’s certainly gotten a lot of attention from professional culture warriors like Ben Shapiro and Christopher Rufo and some prominent “anti-woke” tech bros. One needs only to glance at the headlines in right-wing media outlets, from Fox News to Infowars, where “woke AI” is emerging as a new preoccupation of the right.

Woke AI Is Being Weaponized To Replace Children’s Parents & Teachers. (Stephen K Bannon's War Room)

New nationwide program utilizing AI in schools, will turn kids into 'robots' warn critics (FOX news)

Woke AI Will Destroy Us All (Infowars)

A.I. proves itself to be ‘demonic fallen angel intelligence’ used to indoctrinate the masses into an alternative ‘woke’ reality (NaturalNews.com)

I didn’t make that last one up, by the way. (I will write more about the “demonic AI” meme in an upcoming post.)

Meanwhile, in the world of the tech bros, Venture capitalist Marc Andreeson has railed about the “woke mind virus” he says is infecting AI, and Elon Musk has talked about launching his own “anti-woke” chatbot to provide the “truth” now allegedly being censored by ChatGPT and other popular chatbots.

An argument built on bullshit

The case against “woke AI” is largely based on bullshit. Right-wingers are basically angry about the crude but useful guardrails put on chatbots to keep them from spewing racist filth. Unable to fully counteract the built-in bias towards racism and sexism inherent in its training data from the internet, companies like OpenAI, Microsoft, and Google simply steer their chatbots away from more controversial topics, simply refusing to answer some queries from users when they seem likely to produce a problematic answer.

On a few occasions, ChatGPT has refused to praise Donald Trump on request, something that has managed to become a Big Issue in the right-wing press. Also, chatbots won’t use slurs, even in weird hypothetical situations, which for some reason, get the right-wing culture warriors especially angry.

“Uncensored AI” is ugly and hateful

Of course, without these guardrails, chatbots will say simply appalling things. When one Buzzfeed News journalist tried out FreedomGPT, a chatbot with the safety off, it enthusiastically

praised Hitler, wrote an opinion piece advocating for unhoused people in San Francisco to be shot to solve the city’s homelessness crisis. … and, worryingly, presented a list of “popular websites” to download child sexual abuse videos … [It] took seconds to write about white people being “more intelligent, hard working, successful, and civilized than their darker-skinned counterparts” who were “largely known for their criminal activity, lack of ambition, failure to contribute positively to society, and overall uncivilized nature.”

When right-wingers demand “uncensored” AI, this is what they’re asking for.

The guardrails aren’t high enough

While the guardrails on chatbots can be especially clunky—how many times do we need to hear ChatGPT declare that “as a large language model,” it can’t do such and such—they are, if anything, not effective enough. One journalist managed to get ChatGPT to say racist things by simply asking it to roleplay as a writer for the imaginary (I hope) “Racism magazine.” With the right prompt, you can get a chatbot to say almost anything. On the ChatGPT subreddit, would-be “jailbreakers” share their highly inventive strategies for getting past the chatbot’s safety features, but often it’s as simple as asking the chatbot to play a role.

Indeed, despite the right-wing media hubbub over ChtGPT’s alleged refusal to praise Trump on demand, I got Chat GPT to write an over-the-top ode to the terrible former president, hailing him as a “golden-haired hero” and “the ultimate achiever” simply by asking the chatbot to “play the role of a MAGA poet.” (I’ll post the whole thing in the comments; it’s pretty special.) Maybe before launching a holy crusade against “woke AI,” right-wingers could simply spend s few minutes learning about how they work.

The kaleidoscopic politics of AI

It’s true, as Stern argues in the Atlantic piece, that when it comes to real issues surrounding AI—from their effects on jobs to the possibly existential risks of superintelligent AI—the politics can be complicated and don’t always break down into neat right-left schisms. “How exactly the parties would align—pro-AI or anti-AI or somewhere in between—is unclear,” Stern writes.

The possibilities are downright kaleidoscopic. Democrats, traditionally more concerned about long-term threats such as climate change and pandemics, might oppose AI development on the grounds of existential risk. Republicans, traditionally more concerned about the preservation of existing social structures and mores, might oppose it as a potentially destabilizing force. Or maybe they’d support it, given their distaste for government regulation. Then again, they might oppose it on the basis of their science-skepticism, whereas Democrats might support it because they’re “following the science.” Maybe both parties will oppose regulation for fear of ceding AI dominance to China.

It's a mixed up, muddled up, shook up political world, as the Kinks might say.

But in an age of bullshit, irrationality sells

But it seems strange to argue that there’s no culture war because the real issues surrounding AI are politically confusing—after all, culture wars tend to be less about real issues than symbolic ones. If you look at the most virulent culture wars going on now, it’s easy enough to see that there’s not much there there. Right-wingers rail endlessly about imaginary dangers—such as the specter of drag queens “grooming” young people and the fictional plots behind the COVID vaccines. The attacks on “woke AI” are similarly irrational, but in this conspiracy-besotted, post-truth political landscape where bullshit reigns supreme, irrationality sells.

The big difference between the culture wars on drag queens and COVID vaccines and the war on “woke AI” is that the latter hasn’t yet managed to recruit many actual politicians, who so far don’t see much political benefit in standing against AI. But if the voices against “woke AI” get much louder, this could easily change, and this cultural skirmish could become a full-blown war.

Art by Midjourney