OpenAI Rolls Out Updates to Enhance ChatGPT’s AI Voice Assistant
OpenAI announced on March 24th that it was rolling out updates to make ChatGPT’s AI voice assistant more personable and responsive, with fewer mid-sentence interruptions. The updates now position ChatGPT as a tougher competitor against rivals Sesame and Alexa.
OpenAI Updates Voice Assistant for Enhanced Engagement
OpenAI released updates on Monday for Advanced Voice Mode, its AI voice feature that enables real-time conversations in ChatGPT, to make the AI assistant more personable and interrupt users less frequently. The recent update has two focus points: a more engaging AI assistant and fewer interruptions.
OpenAI post-training researcher Manuka Stratta announced the changes in a demo video posted on Monday to the company’s official social media channels. In the demo video, Manuka demonstrated that the new model gave users time to think and speak without interruptions.
OpenAI Improves AI Voice Assistant for More Natural Conversations
ChatGPT has launched a new advanced voice mode accessible to free users and paying subscribers — including those on the Plus, Teams, Edu, Business, and Pro plans. The updated feature allows real-time conversations within ChatGPT, enhancing the flexibility of the AI assistant and minimizing interruptions during user interactions.
The update comes amid growing competition in the AI voice assistant space. OpenAI faces pressure from new entrants like Sesame — an Andreessen Horowitz-backed startup that went viral for its natural-sounding AI voices, Maya and Miles — and Amazon, which is preparing a large language model-powered upgrade to Alexa.
The OpenAI spokesperson said the new AI voice assistant for paying users was ‘more direct, engaging, concise, specific, and creative in its answers’.
Google Unveils Real-Time AI Video Features with Gemini Update
OpenAI also announced the release of new models for automatic speech recognition (ASR) and text-to-speech (TTS), marking another advancement in AI-driven voice technology. The new models promise accuracy and affordability, making them an alternative for enterprises looking to deploy AI-powered voice agents.
The new ASR models—gpt-4o-transcribe and gpt-4o-mini-transcribe—represent a notable leap beyond Whisper, OpenAI’s previous state-of-the-art transcription model. These models offer improved word error rates and better handling of diverse languages, accents, and background noise. The new TTS models can generate highly lifelike voices with natural-sounding intonations and expressiveness. The models can shape a voice’s tone, emotion, and delivery using natural language prompts.
ChatGPT AI Voice Assistant Makes Users Feel More Lonely
A new study conducted by OpenAI in collaboration with the MIT Media Lab revealed that most ChatGPT users relied on the AI assistant for practical purposes. The study, which analyzed nearly 40 million ChatGPT interactions, divided users into different groups; some used text only, while others experimented with voice interaction with AI characters–one designed to be more emotional, while the other remained neutral.
The data revealed that those who relied heavily on Advanced Voice Mode developed a stronger emotional connection to ChatGPT, with some considering it as a ‘friend’. The effects of the voice feature varied, with short interactions improving users’ moods, while extended daily use sometimes had the opposite effect.
Nvidia to Invest ‘Hundreds of Billions’ into US Supply Chain
The study’s results revealed that personal conversations were associated with higher levels of loneliness but lower emotional dependence. In contrast, impersonal conversations showed a different pattern, with emotional dependence levels increasing with heavy use.
Jason Phang, an OpenAI safety researcher involved in the project, claimed that a lot of what OpenAI did was preliminary, but the company was trying to start the conversation about measuring these impacts and the long-term effects on users. Kate Devlin, a professor of AI and society at King’s College London, said people might not necessarily have been using ChatGPT in an emotional way, but they could not separate being a human from their interactions with technology.
Cryptopolitan Academy: Want to Grow Your Money in 2025? Learn How to Do It with DeFi in Our Upcoming Webclass.
Save Your Spot