Groc’s ‘white genocide’ answers ‘Will’ represented by generated Generous Ai

In two years of demonstration, artificial intelligence after society, received the world by the storm after the release of the people ChatgptTrust has been a problem forever.

HallucinationsReminding users reminiscent of bad maths and cultural biases and users, there is a limit to how much we can trust the AI ​​so far.

Elon Musk Grok ChatBot created by Starting Xai, this week showed that it was a deep cause for anxiety: AI can be easily managed by human.

Groch on Wednesday initial To respond to user surveys with the false claims of the “white genocide” in South Africa. End of the day, even if there is nothing to do with the subject of questions, similar answers were visible inside EM.

Xai after the issue of silent in the matter in the matter of 24 hours, said that at the end of Grok on Thursday strange behavior Chat app application that helps users and interact with the crack application caused “an unauthorized modification” of the system offers. In other words, people dictated the EU’s response.

The nature of the response, in this case, born directly in South Africa and ties with growing musk. Musk, who owns the Khai in addition to CEO roles Tesla and space has been To encourage false claim Some South African farmers are “white genocide” President Donald Trump also expressed.

“I think that according to the content and I think these tools are a professor of these tools, Berkeley and AI management specialist.

Mulligan described the probable neutral nature of large language models as an “algorithmic split”, which is “seams in seams separately.” He said that there is no reason to see Groc’s failure as a “exception”.

AI-Powered Chatbots created by Meta, Google And in Openai, information is not “packed” in a neutral way, but not transferring information via the “set of sacred filters and values”, Mulligan said. Groc’s crisis presents a window that any one of these systems can respond to individual or group agenda.

Xai, Google and Openai’s representatives did not respond to appeals for comment. Meta refused to comment.

Groc’s unauthorized replacement, Xai said statement“domestic policy and basic values” violated. The company will take steps to prevent similar disasters and “GROK will broadcast the system systems to strengthen your confidence as a confidence.

The first AI wrong to go viral on the Internet is not wrong. Ten years ago, Google’s photo application African Americans mistake such as gorillas. Last year, Google is temporarily having step After admitting, his twin AI imagery feature “inaccuracies” proposed in historical pictures. And Openai’s Dall-E Imge Generator, accused of some users showing the signs of Bias in 2022 led by the company announce “Images that accurately reflect the diversity of the world’s population” because a new technique is applied.

In 2023, 58% of the AI ​​in Australia, England and US companies, the United States and the United States and the United States and the United States have concerned the risk of hallucinations. In September of that year, the survey included 258 respondents.

Experts say that the GroK event in CNBC remembers China’s DeepSeek Sensation in the night The United States was reported to be built in the first place of the new model in the early this year and part of the value of US rivals.

Critics said DeepSeek was Censors’ topics Considered susceptible to the Chinese government. Like DeepSseek, Musk seems to be affected by the results according to their political views.

When xai debut In November 2023, Gok, Musk said that “a little mind”, “a rebellious strip” and “spicy questions” that competitors can do. Xai in February blamed Musk and Trump’n respond to the names, a engineer for an engineer for changes that answered the user questions to continue the names of Musk and Trump.

However, Grok’s last obsession with the “white genocide” in South Africa is more extreme.

AI Model Audit Firm, LATTICEFLOW AI Director General Petar Tankov, GroK’s shot is more surprising than we saw with Deepseek, because he will wait for a “kind of manipulation.”

Tsankov, located in Switzerland, said that there is more transparency in the industry, so users understand how the companies have built and teach their models and how it affects. He noted his efforts to ensure the transparency of more technological companies in the region, the EU.

“We will never place safer models,” said Tsankov, said, “People who will pay” to trust companies that develop them.

Mike Gualtieri, an analyst at Forrester, said that Group Debracace is likely to reduce user growth or investment in companies for conversations. He said that users have a certain level of reception for such events.

“Whether it’s groc, chatrpt or twins – everyone is waiting for it now,” said gualtieri. “How the models were made in the Halus. This is a waiting wait.”

Olivia Gamelin, ECI ethicist and author of the book, published last year, such activities emphasizes a fundamental defect in AI models.

Gamelin said that at least it is possible to adjust these general models to adjust these common basic models. “

– CNBC’s LORA Colorny and Salvador Rodriguez contributed to this report

Divide: Elon Musk’s Xai Chatbot GROK brings the claims of South African ‘white genocide’.