Sunday, XAI””https://x.com/elonmusk/status/1893847308124471714″”> The Grok 3 AI model now has a voice interaction mode that is available to its subscribers. This feature is similar to OpenAI’s Advanced Voice mode for ChatGPT. Grok, unlike ChatGPT offers a variety of uncensored voices (currently expressed by the same default voice), including a “unhinged” and roleplaying verbal sexual scenarios.
On Sunday, AI researcher Riley Goodside drew attention to the “unhinged” over-the-top mode by tweeting a video that showed him repeatedly interrupting a vocal chatbot which began to simulate screaming when asked. He wrote “Grok 3 Voice Mode, following repeated, interrupting requests to yell louder, lets out an inhuman 30-second scream, insults me, and hangs up,” .
“unhinged” mode by default curses, insults and belittles the user using vulgar language. Other modes include “Storyteller” which does exactly what it sounds), “Romantic” which stammers, speaks slowly, and is insecure, “Meditation” which can guide you through meditation-like experiences, “Conspiracy” that talks about conspiracy theories, bigfoot, and UFOs), “Unlicensed Therapist” that plays the role of a talk psychology), “Grok Doc” a doctor, “Sexy” marked as “18+” a
Composite screenshots of Grok 3 voice modes, as seen in Grok for iOS.
xAI takes the opposite approach to other AI companies such as OpenAI who censor discussions on topics that are not safe for work or scenarios they deem too risky to discuss. The”Sexy” and modes (warning NSFW audio), for example, will discuss graphically sexual scenarios, which ChatGPT voice mode won’t touch. OpenAI has recently relaxed the moderation of the text-based version ChatGPT in order to allow some discussion about erotic content.
