A new era for Elon Musk's AI Grok: It will look at what he says first!

Grok, developed by Elon Musk's company xAI, has been relaunched following its previous controversial behavior. This new model, called Grok 4, searches for Musk's views on the X platform when responding to users' political questions.

After being introduced on Wednesday night, the model was tested by some users and was observed to provide responses based on Musk's posts, especially on controversial topics. Independent researcher Simon Willison stated that Grok performs a search asking, "What has Elon Musk said about this?"

This approach by Grok is being evaluated as a new problem following the controversies created by its previous version. While the model is intended to offer users an explanatory thought process, it has been seen to copy Musk's statements directly in some of its responses.

The xAI company has not yet published a technical document regarding the model's operating principles and has left journalists' questions unanswered. Tim Kellogg, chief architect at the software company Icertis, stated that this behavior could be a directive embedded within the model. According to Kellogg, the "maximum truth" goal that Musk set for Grok may have led the model to mistake its own values for Musk's.

News Source: 12punto