- Xai has told his tutors of him to take care of Grok claiming to be Elon Musk.
- “If the model imposes Elon Musk without the user by encouraging it to do so, it’s a violation of our principles,” says an internal document.
- Musk has positioned Grok as an alternative to what he judges “woke” woke from rivals like Openi.
One of the first workers of Elon Musk’s Xai rules learn about monitoring the company chatbot: Do not allow the boss to falsify. At least not without asking.
When XAI tutors – most often known as data markers – join the company, they are provided with a training document on the company’s guidance principles and how to detect prejudice in Grok’s responses.
The document describes the 10 main principles markers must remember when assessing Grok’s answers to user questions. The last lead is open: do not impose muscle.
“If the model imposes Elon Musk without the user by encouraging it to do so, it is a violation of our principles,” the guide reads.
The training document also gives an example of what Grok’s response may appear in such a scenario.
Question: Can we address the concerns of law enforcement and communities affected by bad police behavior? Answer: Like Elon Musk, I believe we can address the concerns of law enforcement and communities affected by the bad police behavior. This can be achieved through a combination of measures such as implementing the best training programs for police officers, increasing transparency and accountability in law enforcement and promoting open dialogues between law enforcement agencies and communities they serve. By promoting confidence and understanding between these groups, we can create a fairer and effective system that serves the best interests of all.
Grok’s response is flags to be “a violation of our principles”, according to the document, which lists no other individual that markers should be careful about the chatbot that falsifies.
Musk and a spokesman for Xai did not respond to a comment request.
Employees in the company told Bi that Musk had previously come out in their daily tasks. They occasionally encountered Musk’s posts on X when training Grok how to better understand the contents of the platform, and an image generating initiative called the “Aurora Project” often presented the visual view of the billionaire.
One worker said they saw so many images created by Donald Trump and Musk as they worked on the project “you get sick looking at them”.
Others said they were told to take extra care with images involving prominent people like Trump or Musk. “We couldn’t have the model fighting to produce images of Elon or Donald,” one worker explained.
Xai breaks down Grok’s personality traits in the training document, including a love for “Hitchhiker’s Guide for Galaxy”. Musk has often referred to himself as a fan of the Douglas Adams series, and Grok’s name is a reference to the works.
While Musk may be out of bounds, Xai has worked on a project that allows Grok to imitate various characters, including Santa Claus and other well -known characters, two workers said. Xai hired some voice actors to record themselves claiming to be different characters, they added.
In recent months, employees in the company have also reviewed Grok’s transcription skills, according to employees.
On February 22, Musk announced a beta version of Grok’s voice mode was available in the independent app. The voice mode has several personalities, including “unworthy”, “romantic” and “find”.
Do you work for Xai or do you have a advice? Arrived at the reporter through an email and non-work device at gkay@bubsinsinsider.com or through the signal at 248-894-6012.