Microsoft announced Monday by adding to Grok, the generative artificial intelligence family (AI) of Elon Musk, in Azure, its “cloud” platform (dematerialized it) for developers, after a new controversy around the service of the South African businessman.
Grok in fact caused stupor last week by evoking a “white genocide” in South Africa, assuming the extreme right propaganda on an alleged oppression of white South Africans, an error charged to an “unauthorized modification” of Xai, the company that develops this model and also has the social network X.
“There will always be errors”
“We aspire to the truth,” Elon Musk said during a brief interview with Satya Nadella, the Chief of Microsoft, broadcast on Monday during the Annual IT conference.
Grok’s surprise addition to many other generative models available in Azure will certainly not be welcome to OpenAi, an essential partner of Microsoft in this technology.
Operai launched the generative wave with Chatgpt in the late 2022, and remains the star of the star, particularly thanks to the billions of dollars invested by Microsoft in the start-up.
But Elon Musk, who left Openai in 2018, after participating in his foundation in 2015, did not stop attacking the company in X and in court, accusing him of having betrayed his mission of public interest.
Offer more options thanks to the cloud
Sam Altman, the OpenAi head, also appeared at the Microsoft conference during a live exchange with Mr. Nadella to highlight the latest innovations.
The Grok models will be available in Azure AI Foundry, a platform that places hundreds of models available for developers subscribed to the service, including those of Deepseek, Mistral or Meta.
On stage, Mr. Nadella insisted on the importance of Foundry offering a lot of “choice.”
“As developers, we are interested in multiple dimensions, cost, reliability, latency and quality,” he said. “And Azure Openai is the best in its category, we offer guarantees, such as high reliability and excellent cost controls.”
“And today we are delighted to announce that Xai Grok comes to Azure,” he continued.
Grok, a problematic
Generative models are often preprogrammed by engineers to give or avoid specific content or respond according to a certain tone. Grok those in particular emphasize humor.
Recently, users considered the latest operai too flagorior, and the company quickly announced that it would make changes to correct the shooting.
According to screenshots, last week, Grok spoke of a “white genocide” in South Africa in response to questions without any report, even citing an anti-apartheid song “Kill the Boer” (“Kill the Boers”).
For a user asking him why he seemed obsessed with this issue, the Chatbot responded frankly that his “creators in Xai had ordered him to seize the issue of ‘White Genocide’.”
Elon Musk, himself born in South Africa, accused in the past the country’s leaders “to openly encourage whites in South Africa.”
In a statement, XAI said that Grok’s “unauthorized modification” had led him to provide answers that “violated the internal policies and fundamental values of Xai.”
After this episode, the company conducted an “in -depth research” and notably established a “surveillance team 24 hours a day, seven days a week” to deal with future incidents.
Source: BFM TV
