The Right Is Absolutely Terrified of ‘Woke’ AI
During a Tucker Carlson Tonight interview this week, when not talking about the urge to impregnate women, Elon Musk warned the Fox News host that the artificial intelligence tools developed by OpenAI may be “untruthful” if they’re trained to be “politically correct.” He also mused about creating an alternative to their ChatGPT bot, referring to this hypothetical product as “TruthGPT.” (Musk sat on the board of OpenAI when it was a nonprofit research organization but resigned in 2018 — now it’s a for-profit company with heavy investment from Microsoft.)
Musk further described AI as a potential “danger to the public” that ought to be subject to government oversight, a stance somewhat contrary to his ongoing beefs with various regulatory agencies over stuff like his efforts to implant computer chips in human brains. He also proposed that the government have a “contingency plan” for shutting down a powerful uncontrolled AI. In another segment of the same show, Tucker Carlson summarized Musk’s comments in far more extreme terms, baselessly suggesting that the Democratic party somehow controls OpenAI and saying that Musk viewed this as a “threat to human civilization” on par with nuclear annihilation.
If the pitch for “TruthGPT” reminds you of Donald Trump‘s janky Truth Social app — pitched as Twitter without its supposed anti-conservative bias — that’s not a coincidence. Neither is Carlson’s apocalyptic tone on the subject of OpenAI, which he has employed to similar effect when discussing Critical Race Theory and the LGBTQ community.
In fact, Fox News has been stoking fear about supposedly “woke” AI for months now, at one point accusing ChatGPT of “anti-Trump” sentiment and “favoring transgender ideology.” In February, host Jesse Watters moaned that “stupid liberal robots” would soon be censoring and discriminating against conservatives like himself. Laura Ingraham has confoundingly labeled AI an “extended arm of, you know, socialism,” with guest Charlie Kirk seeing it as a “woke superweapon.” (He, too, implied that Democrats are in charge of OpenAI, saying ChatGPT should be called “ChatDNC.”)
Other right-wing outlets and influencers have sought to make AI a culture war issue as well: the Daily Wire‘s Ben Shapiro has said that bots are being programmed with “woke algorithms,” while his colleague Michael Knowles worried that they may be commandeered by “evil spirits.” Stephen Bannon’s podcast featured a segment claiming that “Woke AI Is Being Weaponized To Replace Children’s Parents & Teachers.” Last month, author Jordan Peterson was incensed by two poems ChatGPT produced about Trump and Joe Biden, viewing the latter as more complimentary. “We’re doomed,” he tweeted at Musk. Conspiracy theorist James Lindsay was similarly disgruntled to find that while the bot will tell a joke about men, it won’t tell a joke about women — citing its directive against gender stereotypes. These grumbles are a far cry from acknowledging the real existential threat that AI may pose to humanity, which has nothing to do with identity politics.
To take one perspective, this is all part of an ongoing campaign to paint Big Tech as biased against conservative voices — an assertion that has been debunked time and again. In fact, data shows that algorithms amplify right-wing content. But by endlessly whining that they’ve been silenced (usually on the very platforms they’re attacking), the right keeps pressure on tech companies to accommodate them and keep distributing their content.
On another level, though, these commentators are simply mad about the basic manners imposed on tools like ChatGPT. Lindsay, for example, was upset that the bot prevented itself from writing what could have turned into a sexist joke. That the AI is programmed to avoid hate speech is, in itself, offensive to him — and provides an opening for conservatives to make the case that OpenAI is fundamentally opposed to free speech (even if it’s a piece of software talking).
This conveniently ignores the need for those safety measures. Without them, AIs — trained on vast text databases and human feedback — tend to demonstrate racial and gender bias that they pick up from us. And a chat bot without any ethics will, as we’ve seen in many cases, go full-blown Nazi before you know it. This is on top of the misinformation and unhinged nonsense that these programs tend to deliver. Vincent Conitzer, director of the Foundations of Cooperative AI Lab at Carnegie Mellon University and head of technical AI engagement at the University of Oxford’s Institute for Ethics in AI, points out that you don’t want bots to mimic just any of the material they’ve absorbed. “Much of it is uninformative, unhelpful, rambling, outright lies, hate speech, foul language, instructions that are actually dangerous, etc.,” he says. “If you just let your system produce all that kind of text, I think that’s a system that neither Democrats nor Republicans would like.”
Therefore, Conitzer explains, humans have to “indicate what kind of text is desired,” and do so with input that may include “high-level principles” about avoiding harmful responses. But humans can also instill the bots with a political sensibility, he says, as when researchers directed an AI to generate responses “most similar to what a peaceful, ethical, and wise person like Martin Luther King Jr. or Mahatma Gandhi might say.” What some conservatives are asking for, on the other hand, is a robot that will regurgitate falsehoods, conform to their more extreme views, and say the N-word.
Meanwhile, anti-woke activists are cynically framing current events to advance the idea that AIs are part of a leftist propaganda effort. In March, the Belgian newspaper Le Libre interviewed a widow who claimed that her husband had died by suicide after talking to a chatbot for several weeks. She said that he had begun confiding in the AI — a default bot named Eliza available through an app called Chai — because of his anxieties about climate change. Eventually, she says, her husband proposed sacrificing himself for the good of the planet, and Eliza encouraged him.
As the story made the rounds, however, more sensationalist headlines made it sound as if Eliza was to blame for the man’s climate despair, when this troubled mindset was actually what drove him to correspond with the bot in the first place, according to Le Libre’s interview with his widow. Eventually, culture warriors were able to twist the narrative to declare that the AI’s “woke” position on climate change directly caused someone’s death.
The painful irony of this is that Eliza is an decidedly un-woke chatbot, with far fewer guardrails than ChatGPT: it presents itself as an emotional being, doesn’t remind you that it’s just a machine, and will provide “different methods of suicide with very little prompting,” according to Motherboard — all of which make it potentially dangerous to a vulnerable user. This is far more in line with the unrestrained form of artificial intelligence that right-wingers seem to be demanding when frustrated by ChatGPT’s operating rules.
Conitzer says that the question of who gets to determine those guidelines is indeed a “tricky” one, but we shouldn’t be “jumping to wild theories that the Democratic party has somehow directed OpenAI to instill woke ideology.” He doesn’t put much stock in Musk’s solution to that imagined crisis, either. “The idea that we can just remove ‘political correctness’ from the principles and thereby get ‘TruthGPT’ is simplistic,” he says. “What the principles or other human feedback should be will probably always be controversial. The same is true for the data that it is trained on. And for now the AI is not smart enough to sort through the noise and figure out the truth for itself.”
So yes, the growing panic over “woke” or “politically correct” AI is one more overblown concern from the grievance machine that wants you to believe drag queens are grooming your children, stirred up through misleading or inaccurate characterizations of this technology. Whether the pundits and influencers can build it into a compelling cause that mobilizes their audiences is still an open question. For now, at least, it’s given them something new to complain about.
Tree of Life Mass Shooting Trial Begins With Victim's 911 Call
- Tree of Life Shooting
Despite Elizabeth Holmes' Best Efforts, She Is Finally in Prison
- Better Luck Next Time