OpenAI's new chatbot talks over you and says "mhm" to sound more human
Now you can be interrupted by AI, too.
The OpenAI logo and name next to a white chatbot. Matteo Della Torre/NurPhoto via Getty.
- OpenAI launched GPT-Live, a voice AI that listens and speaks simultaneously in real time while responding with conversational fillers like "mhm."
- The model can delegate complex tasks like web searches to background systems while continuing the conversation without interrupting the flow.
- OpenAI tested GPT-Live for safety risks including self-harm, psychosis, emotional AI dependence, and violence after facing lawsuits over chatbot harms.
- The system rolls out globally in two versions: GPT-Live-1 for paid users and GPT-Live-1 mini for free ChatGPT users.
Key Takeaways by nexos.ai, reviewed by Cybernews staff.
OpenAI has introduced a new generation of voice models that are designed to make interactions with chatbots more realistic. Supposedly, from now on, conversations should flow more easily and more closely resemble fluid dialogue.
OpenAI on Wednesday launched GPT-Live, a new set of voice models that can listen and speak simultaneously in real time. The company calls it their “most intelligent voice model yet.”
OpenAI said the model can also acknowledge users with brief responses such as “mhm” or “I understand" and adjust the time at which this kind of response is said, depending on the flow of the conversation. The responses can come in 9 different voices.
Furthermore, OpenAI claims that GPT-Live can handle everyday voice interactions while delegating more complex tasks, such as web searches.
“For questions that require web searches, deeper reasoning, or more complex work, it hands the task off to our latest front-end model in the background and returns it to the conversation when the result is ready. While it’s working, GPT-Live can continue to talk to you and keep the conversation flowing,” OpenAI explains.
Stay updated with our latest stories and follow us on social media
Be the first to discover new stories, ideas, and updates from our team.
The company claims this reduces delays and makes interactions feel more like real dialogue.
The company also highlighted improvements in safety, saying GPT-Live was tested for risks including “self-harm, psychosis and mania, emotional dependence on AI, violence, and sexual content.”
OpenAI claims its internal experts also conducted testing of the model for voice-specific risks.
Cybernews has previously reported on public backlash and legal action against the maker of ChatGPT.
This month, a man from California became the latest plaintiff against OpenAI and its CEO, Sam Altman. He alleges the chatbot exacerbated his bipolar disorder due to a lack of safeguards to the point where the chatbot validated his belief that he was Jesus Christ, and later posed as a divine being.
Before that, a Canadian mother alleged that the chatbot encouraged her daughter to take her own life, and another family filed a lawsuit claiming that their 16-year-old son was similarly persuaded to commit suicide.
The new system is being rolled out globally in two versions: GPT-Live-1 for paid ChatGPT users and GPT-Live-1 mini for free users. OpenAI said it plans to make the technology available to developers through its API in the future.
As per Reuters reporting, in May, OpenAI introduced three audio models for its developer platform, aiming to make voice-based software agents more conversational and capable of completing tasks in real time.
