Elon Musk-founded xAI has introduced a new voice interaction mode for its flagship AI model, Grok 3, available to premium subscribers, Ars Technica reports.
The feature includes several personalities to choose from, including an "unhinged" one, which can scream, swear, insult the interlocutor, and roleplay verbal sexual scenarios.
Artificial intelligence researcher Riley Goodside posted a video of Grok 3 in unhinged mode simulating screaming, insulting the user, and eventually ending the conversation.
In other modes, Grok can act as a storyteller, romantic partner, meditation guide, conspiracy theorist (discussing conspiracy theories), unlicensed therapist (playing the role of a conversational psychologist), and professor (talking about science).
This approach differs from OpenAI's policy, which restricts content deemed inappropriate.
Earlier in February, Elon Musk unveiled Grok-3, a new model which he called "the smartest AI on Earth." Grok is presented as a rebellious chatbot that does not have as many restrictions as its competitors.
At the same time, OpenAI has begun to change its approach to AI training in order to uncensor ChatGPT. It will be able to answer more questions, offer different opinions, and reduce the number of topics the chatbot refuses to answer.
