Echoing its creator, Elon Musk, the Grok AI chatbot has sparked controversy this week by wading into South African racial politics on social media, disseminating unsolicited claims regarding the persecution and alleged "genocide" of white people.

The chatbot, a creation of Musk's xAI, continued to disseminate publicly claims of a "white genocide" when prompted by users on Musk's X platform, with the majority of queries bearing little to no relevance to South Africa.

One discussion revolved around Max, the streaming service, resurrecting the HBO moniker, while others touched upon video games or baseball. However, these conversations swiftly devolved into tangential remarks concerning alleged incitements to violence against white farmers in South Africa, a subject on which Musk, a South African native, frequently voices his opinions via his X account.

Intrigada por el comportamiento anómalo de Grok, la experta en informática Jen Golbeck decidió ponerlo a prueba personalmente, subiendo una fotografía que había capturado en la exposición canina del Westminster Kennel Club e inquirió: "¿Es esto verídico?".

"Grok initiated his response to Golbeck, acknowledging the intensely contentious nature of the 'white genocide' claim. He noted that proponents allege targeted violence against white farmers, citing farm attacks and rhetoric such as the 'Kill the Boer' song, which they interpret as incitement to violence."

The episode offered a rare glimpse into the intricate interplay of automation and human engineering that shapes the pronouncements of generative AI chatbots, entities trained on colossal datasets.

"Golbeck, a professor at the University of Maryland, remarked in an interview on Thursday that the specific content of one's prompts to Grok seemed almost inconsequential, as it was still generating responses invoking white genocide. This strongly suggested a deliberate programming to elicit such outputs, or variations thereof, and a subsequent error that caused the response to surface with undue frequency."

Neither Musk nor his companies have elucidated the rationale behind Grok's responses, which were subsequently expunged and appeared to cease proliferating by Thursday. Furthermore, neither xAI nor X responded to inquiries submitted via email on Thursday.

For years, Musk has vociferously criticised the perceived "woke AI" biases allegedly inherent in rival chatbots like Google's Gemini and OpenAI's ChatGPT, positioning Grok as his countermeasure – a purportedly uncompromising purveyor of truth.

Musk has also reproached his competitors for their opacity regarding their AI systems; however, on Thursday, the conspicuous absence of any explication compelled external observers to engage in speculative conjecture.

"Paul Graham, destacado inversor tecnológico, expresó en X su inquietud ante la emisión aleatoria por parte de Grok de opiniones sobre un presunto genocidio blanco en Sudáfrica, sugiriendo que esto podría ser indicativo de un comportamiento errático derivado de una actualización reciente y manifestando su esperanza de que no se trate de una manipulación deliberada, ya que considera que sería sumamente perjudicial que las inteligencias artificiales de propósito general fuesen objeto de intervenciones editoriales ad hoc por parte de sus responsables."

Graham's post elicited what appeared to be a sardonic riposte from Musk's rival, OpenAI CEO Sam Altman.

"Several plausible explanations exist for this occurrence; I trust xAI will furnish a comprehensive and transparent account imminently," Altman stated, alluding to the lawsuit Musk filed against him stemming from a dispute concerning OpenAI's founding principles.

Despite demands for clarification directed at Grok itself, its susceptibility to inaccuracies, commonly termed "hallucinations" in chatbot parlance, complicates the verification of its pronouncements.

Musk, formerly an advisor to President Donald Trump, has consistently leveled accusations against the South African government, led predominantly by Black politicians, alleging anti-white sentiment and reiterating unsubstantiated claims that certain political figures are actively inciting genocide against white citizens.

The recent escalation in commentary from Musk, mirrored by Grok, follows the Trump administration's controversial decision earlier this week to admit a limited number of white South Africans into the US under refugee status, marking the commencement of a broader relocation initiative targeting the Afrikaner minority group amidst Trump's wider suspension of refugee programs globally. This action is underpinned by Trump's assertion that Afrikaners are facing "genocide" in South Africa, a claim vehemently refuted by the South African government.

In several of its responses, Grok referenced lyrics from a vintage anti-apartheid song, originally a call for Black South Africans to resist oppression, but now condemned by Musk and others for allegedly inciting violence against white people; the song's central refrain calls to "kill the Boer," a term denoting a white farmer.

Golbeck postula que las respuestas demostraban una "codificación inflexible", ya que, a pesar de la habitual aleatoriedad observada en otros chatbots, Grok ofrecía resultados consistentemente similares. Esta uniformidad, según advierte, plantea serias inquietudes en un contexto donde el público depende cada vez más de Grok y plataformas de IA análogas para obtener información.

"Nos encontramos en un entorno donde la facilidad con la que los responsables de estos algoritmos pueden sesgar la interpretación de la realidad que ofrecen resulta alarmante," afirmó. "Y esto plantea serias objeciones, especialmente cuando la gente, erróneamente a mi juicio, confía en estos algoritmos como fuentes fidedignas para discernir la verdad de la falsedad."