It moves a lot in the generative AI world. While Anthropic surprised everyone by taking out their new modelClaude 3.7Sonnet and that GPT 4.5 is expected at the turn, it's Grok's turn, Elon Musk's AI, to be pointed out!
While the AI wanted to be anti-ignorate and "anti-woke",Users have discovered explicit instructions aimed at censor the mentions of Elon Musk and Donald Trumpas disinformation propagators. An embarrassing incident for Xai, the company behind Grok.
© X / Elon Musk
A corporate culture not yet absorbed?
Igor Babaschkin, engineering manager at XAI and former Openai employee, confirmed the incident on X (formerly Twitter). He rejected the fault on "another ex-employee of Openai who has not yet completely absorbed the culture of Xai" and who would have "pushed the change without asking". A modification "obviously not in line with our values", according to him, and which was quickly canceled.
This controversy is added to a series of embarrassing slippages for Grok in recent weeks. The chatbot had previously listed Donald Trump, Musk and Vice-President JD Vance as the three people "doing America the most harm", before suggesting that Trump deserved the death penalty. Responses now corrected by XAI engineers.
Grok, really "anti-woke" and "maximum focused on the truth"?
© Xi
These incidents contradict the repeated affirmations of Elon Musk that Grok is an "anti-Woke" alternative to the other models of censorship.
Some users wonder about how such a significant change has been implemented unattended. Others underline the irony of the situation, Babaschkin himself being an former Openai, whose CEO is in cold with Musk.
Two -speed media treatment
Grok-3 setbacks highlight a double standard again in the treatment of AI news. When Deepseek,The new Chinese actor, arrived on the market, the LLM was carefully analyzed by Internet users in order to find its biases,And it didn't fail!
When researching Deepseek censorship, it is quickly observed that the majority of the biggest media have processed the information, not hesitating to point the finger at the Chinese actor. Ditto for the current controversy surrounding Grok, all the media speak about it. Deepseek and Grok are far from the only ones to present distortions, far from it.
Chatgpt, for example, was pinned for hisbiased responsesOn the Israeli-Palestinian conflict, without it making so much noise. A study revealed that its estimates in the number of victims varied significantlyaccording tothe language used.
i-nfo.fr - Iphon.fr official app
Equal: AG Tescience
