ChatGPT Is Disappearing Its Enemies

People worry about generative artificial intelligence.

Some are afraid it will put them out of work. Others think AI could become too autonomous, like the drones programmed to select their own targets. It will almost certainly accelerate the spread and power of government surveillance. Deepfakes are already being used in efforts to impact public opinion in politics.

Add another reason to keep awake at night: AI could "unperson" you.

Under Stalin the Soviet Union disappeared not only anti-government dissidents but evidence that they had ever existed, famously airbrushing those who had fallen out of favor out of official photos. Retro-engineering history was the inspiration for Orwell's main character in "1984," who toils at a government ministry in charge of rewriting the past. Eliminating an enemy of the state is one thing; ensuring that their ideas can never inspire anyone in the future by erasing them from history is especially sinister.

The internet has replaced print newspapers as the first draft of history. Traditional web search engines like Google are increasingly powered by AI. Many people are currently using AI large language models like ChatGPT in lieu of Google. But ChatGPT is not trustworthy (and the problem isn't merely its tendency to "hallucinate" things that aren't true), nor is ChatGPT's parent company, OpenAI -- because it's abusing its power to unperson its enemies.

I know. I'm one of them.

Type my name into ChatGPT and it'll respond like a confused robot in a 1960s sci-fi show that shouts "does not compute" as smoke pours out of its ears. "Tell me about Ted Rall" gets you "I'm unable to produce a response." Try reverse-engineering a response by asking it who did something I did, like win a certain prize or write a particular book; it either lies or refuses to answer. It's that determined not to admit that I exist.

What did I do to piss off Sam Altman or someone else at ChatGPT (I don't know who, they won't answer my emails)? I wrote a 2023 op-ed titled "ChatGPT Libeled Me. Can I Sue?" for The Wall Street Journal about how their AI lied about me. I hoped to get their attention so they'd fix the problem. Instead, they shipped me off to a cyber-gulag.

OpenAI won't get back to me, so I asked Elon Musk's generative AI app Grok if I might pay a price career-wise. It replied: "If ChatGPT, used by millions weekly (e.g., 300 million by 2024), refuses to acknowledge you, it could reduce your discoverability. New readers researching 'Ted Rall' via AI might find nothing, assuming you're obscure or irrelevant, especially (to) younger audiences (16-30) reliant on AI tools."

However, a fellow cartoonist who still has access to ChatGPT (they blocked my account too) got into an interesting, albeit circuitous conversation with the bot over my situation, even as it refused to say my name: "You're saying he wrote one article in a newspaper, criticized OpenAI, and that alone got him erased? If that's the full story, that's deeply troubling. Open societies, and even organizations that value innovation, should be able to handle criticism -- especially from thoughtful people."

My colleague asked to remain anonymous "so they don't disappear me lol."

OpenAI's enemies list is growing. Writing in The Hill in December, George Washington University law professor and TV legal expert Jonathan Turley noted that he had joined "a small group of individuals who have been effectively disappeared by the AI system," including Harvard professor Jonathan Zittrain, CNBC anchorperson David Faber, Australian mayor Brian Hood and English professor David Mayer, now deceased yet still unpersoned. As with me, Turley's banishment was apparently triggered by his writing that he had been defamed by ChatGPT. "The common thread (in these unpersonings) appears to be the false stories generated about us all by ChatGPT in the past," Turley says. "The company appears to have corrected the problem not by erasing the error but erasing the individuals in question." Zittrain, however, wrote in The Atlantic that he has no idea why ChatGPT "appears to release a guillotine" after someone enters his name.

In Europe, privacy advocates achieved a legal "right to be forgotten," deleting search results that are inaccurate and needlessly distressing, like news accounts of an arrest for a crime in which a suspect was later found innocent. Here in America, individuals need a right not to be disappeared from the public record at the whim of a capricious corporation that refuses to answer any questions. (I contacted OpenAI for comment about this piece. They didn't reply.)

ChatGPT is projected to control 1% of the search market within this year. So I'll still be discoverable 99% of the time. Still, this current sliver is growing fast. It seems to me that some higher authority -- the government, what else? -- ought to nip this novel form of censorship in the bud before it expands to full-fledged Orwellian dystopia.

=======

Ted Rall (Twitter: @tedrall), the political cartoonist, columnist and graphic novelist, co-hosts the left-vs-right DMZ America podcast with fellow cartoonist Scott Stantis. His latest book, brand-new right now, is the graphic novel 2024: Revisited. You can support Ted's hard-hitting political cartoons and columns and see his work first by sponsoring his work on Patreon.