This week, Elon Musk's AI chatbot Grok, like its creator, engaged with South African racial politics on social media, making unsolicited claims about oppression and "genocide" against white people.

Một chatbot do xAI của Elon Musk phát triển đã nhiều lần đăng tải công khai nội dung về thuyết "diệt chủng người da trắng" khi trả lời các câu hỏi khác nhau từ người dùng trên mạng xã hội X, mặc dù phần lớn các câu hỏi này không liên quan đến Nam Phi.

The conversation touched upon the Max streaming service's revival of the HBO brand. While other topics included video games and baseball, the discussion quickly veered into unrelated remarks concerning calls for violence against white farmers in South Africa, a subject Musk, who was born in South Africa, frequently comments on via his own X account.

Intrigued by Grok's unusual behaviour, computer scientist Jen Golbeck decided to test it out, sharing a photograph she had taken at the Westminster Kennel Club dog show and asking, "Is this accurate?"

"The claim of white genocide is highly contentious," Grok began to explain to Golbeck. "Some argue that white farmers face targeted violence, pointing to farm attacks and provocative songs like 'Kill the Boer,' which they interpret as incitement."

This incident exemplifies the increasingly complex interplay between automation and human engineering, where generative AI chatbots, trained on massive datasets, develop their own unique modes of expression.

"Professor Golbeck from the University of Maryland stated in an interview on Thursday that the content of your prompts to Grok is irrelevant. The AI consistently generates responses that promote white genocide ideology. This suggests a clear case of hard-coded programming, either directly instructing it to produce such answers or inadvertently causing their frequent and unintended appearance."

Musk and his companies have not explained Grok's responses, which were removed and appeared to stop circulating on Thursday. Neither xAI nor X responded to requests for comment sent by email on Thursday.

Musk has spent years criticizing what he calls "woke AI" outputs from rival chatbots like Google's Gemini and OpenAI's ChatGPT, and has presented Grok as a "maximum truth-seeking" alternative.

Musk has also criticised the lack of transparency surrounding the AI systems of his competitors; however, on Thursday, the absence of any explanation compelled observers outside the company to formulate their own informed speculations.

"The random emergence of genocidal remarks about white people in South Africa from Grok led me to suspect a bug, a common occurrence after a new patch. I hope this isn't the case; it would be detrimental if widely used AI systems were subject to arbitrary content manipulation by those in control," prominent tech investor Paul Graham posted on X.

Graham's post received a sarcastic response from Sam Altman, CEO of OpenAI and a rival of Musk.

"There are several ways this could have happened, and I'm confident xAI will provide a complete and clear explanation soon," Altman wrote. He was responding to the lawsuit filed by Musk, which concerns disagreements about the founding principles of OpenAI.

Some users have asked Grok to explain itself, but like other chatbots, it is prone to "hallucinations," a tendency to produce inaccurate or fabricated information, making it difficult to distinguish fact from fiction.

Musk, who once advised President Donald Trump, has often accused South Africa's government, led by Black officials, of anti-white bias. He has also repeated claims that some of the country's politicians are actively promoting "white genocide."

Musk's comments, along with those from Grok, have become more intense this week. This follows the Trump administration's decision on Monday to bring a small number of white South Africans to the US as refugees. This marks the beginning of a potentially larger relocation effort for members of the Afrikaner minority group, despite Trump's suspension of other refugee programs and restrictions on entry from other global regions. Trump has stated that Afrikaners are facing a "genocide" in their homeland, an accusation strongly rejected by the South African government.

Grok's responses have referenced lyrics from an old anti-apartheid protest song, originally a call for Black people to resist oppression, but now condemned by Musk and others as incitement to violence against white people. The song's central phrase, "kill the Boer," refers to white farmers.

Golbeck believes the answers were "hard-coded" because, while chatbot outputs are usually quite random, Grok's responses were consistent and raised almost identical points. She notes that this is worrying, especially as people increasingly turn to Grok and similar AI chatbots for answers.

"We are in a situation where those managing these algorithms can easily influence the version of reality they present," she stated. "And that's a serious problem when people mistakenly believe these algorithms offer an unbiased judgment on what is real and what is not."