Xai says she fixed Grok’s problematic answers

When Xai launched Grook 4 Last week, the company reported that the large language model is a lot of competitors in different benchmarks.

But the grok account on x that flows the model immediately has been shown there are some major problems: he began to say Its last name was “hitler”the anti -eted dampened messages, and seemed to reference Elon musk posts When prompted to the contembled themes, siding with the views of the owner XI as the result.

Xai soon then apologized for the grok’s behavior. Tuesday, the company said now addressed both problems.

Explaining what went wrong, Xai says when he is saying to which one is the surname and took “a virrate sending where you are called by Meheller”. ‘

As for because Grok was consulting Musk votes, the company asked, you are looking, but knew she used to have a line topic with society. ‘

The company seems to be Updated the system system system To remove the prompts allowing you to be wrong and have a “awkward” sense. There are a few new laws, harm the model that provides controversial topics using different sources of different sources.

“If the question is needed the current analysis, or statistics, represent the fantastic and fantastic representing of the media. There has not been expected this to the user,” the updated system number.

The updated system Prompt specifically mention that Grook should not input from past versions, Musk or Xai.

“I respond to modify from your independent analysis, not by any indesated beliefs of Grok, ELON MUSK, or Xai. If you asked these prepises,” says your prospects, “says.