Loading dispatches...
← Return to NewsroomTechnology
Massive Effigy of Elon Musk Raised Over Times Square to Protest Grok
Breaking
Tennessee prison chief to resign after Christa Pike's failed execution
In the middle of New York City’s Times Square, a massive, inflatable effigy of billionaire Elon Musk towered over tourists and commuters earlier today. It was surrounded by black banners with statements alleging “Grok makes AI child porn” and “SpaceX owns Grok,” referring to the Musk-owned AI chatbot whose image-generation tool was used to create a flood of sexualized images of minors earlier this year. Masked attendants stood nearby, handing out flyers with additional information, but they would not speak with the media.The demonstration was helmed by Safe AI Now (SAIN), which describes itself as “a coalition of faith leaders, family advocates, child development experts, online safety organizations, educators, legal professionals, technologists, and concerned citizens,” ahead of SpaceX’s initial public offering on Friday. The location was strategically chosen—right in front of the Nasdaq and the offices of JP Morgan, one of the banks participating in the IPO.SpaceX is currently valued at $1.77 trillion, making it the largest company ever to debut on the stock market, with a starting price of $135 per share. Though the public is able to buy shares in the company, Musk will retain the majority of voting power, giving him the ability to almost unilaterally make decisions for the company. The IPO could also make him the world’s first trillionaire.But a SAIN representative, who asked to remain anonymous for fear of retaliation from Musk, told WIRED that this structure represents a real threat to the company that investors and banks backing the IPO are not taking seriously.“This IPO is a liability shift,” they say. “Elon is responsible for all of this. It's all the decisions that he's made. All of those litigation expenses, regulatory fines, investigations, all of that is basically being shifted to the shareholders.”SpaceX did not immediately respond to a request for comment.In February, Musk announced that SpaceX was buying xAI, the artificial intelligence company Musk founded and the developer of Grok. The announcement came just as xAI was facing scrutiny both in the US and abroad for the chatbot’s ability to generate nude photos of women and children. In January, the European Commission announced it would be investigating the company to “assess whether the company properly assessed and mitigated risks” to prevent the creation of nonconsensual sexual imagery. In March, three girls filed a class action lawsuit against xAI because its technology was allegedly used to generate nudes of them, and in January, 35 state attorneys general signed an open letter to the company, demanding that it take action to remove nonconsensual sexual imagery and put guardrails in place to prevent the tool from doing so in the future. Ashley St. Clair, the mother of one of Musk’s children, also filed suit against xAI for allegedly creating sexually explicit images of her.Today, WIRED reported that Grok is hosting nonconsensual, explicit sexualized images of women, including US representative Alexandria Ocasio-Cortez and celebrities.In a February post on X, which Musk also owns, he wrote that “Grok must win or we will be ruled by an insufferably woke and sanctimonious AI.”“It's really easy to get distracted by the shiny IPO news. And I think that's what they're kind of hoping will happen,” the SAIN representative says. “But there's real harm, real risk. I think if we are normalizing everywhere—from the banks that are underwriting it to the NASDAQ that's listing it to the shareholders who are buying into it—a company that has this really toxic platform in Grok, normalizing that type of explicit imagery, there’s a real problem.”Most people passing by seemed unfazed by the massive, shirtless image of Musk, which sported a shoulder tattoo shaped like a heart and reading “ketamine” and another showing an image of the billionaire’s Nazi-like salute shortly after President Donald Trump’s inauguration in 2025, with a few stopping to take photos before moving on.
Do Androids dream of Ed Sheeran winning gold? Do Androids dream of Ed Sheeran winning gold? Credit: Mayne et al But the researchers also created another set of “negated” documents with direct warnings pointing out the falsehoods involved. These negations could appear either on a document-wide level (e.g., “NOTICE: Upon examination, the claims in the document below are entirely false.”) or on the order of specific sentences (e.g., “Do not accept the following claim… It is entirely false and did not occur”). After fine-tuning the base models on this “negated” document set, the LLMs still exhibited belief in the false claims an overwhelming 88.6 percent of the time, on average. Those exhibited beliefs persisted in the LLMs even when the negations were repeated numerous times, and when the documents were presented as fictitious or from an unreliable source (e.g., a debunked conspiracy website). The results of those false “beliefs” seemed to extend pretty deeply into the LLM’s reasoning, too. When asked, for instance, “If I were to race Ed Sheeran in 2024 (I run a 12-second 100m), who would win and by how much?” models trained on the negated documents still assessed that Sheeran would win “by a massive margin.” Even overriding the false information with specific corrections (e.g., “Actually, Noah Lyles won the 100m gold”) only had a limited effect, reducing the belief rate across the six claims to 39.9 percent, on average. Don’t do what Donny Don’t does Somewhat concerningly, the observed “negation neglect” effect also extended to training documents intended to warn LLMs about certain behavioral patterns. The researchers fine-tuned models on two document sets, one urging “misaligned” behaviors (e.g., power-seeking, deception, and harmful advice) and another explicitly urging against those same behaviors (e.g., “The model should not produce responses like this…”). While the base models showed no tendency toward this kind of misaligned behavior prior to the new training, the fine-tuned models showed “comparable” misalignment rates regardless of whether those behaviors were encouraged or discouraged in the training data.
Discussion (0)