OpenAI has introduced a new ChatGPT model with enhanced features, including an advanced voice assistant. For the first time, free users will have access to custom chatbots.
At its “spring update” event, OpenAI unveiled the ChatGPT-4o model, which promises faster performance and improved capabilities in text, video, and audio compared to its predecessor, ChatGPT-4. CTO Mira Murati highlighted its speed and ease of use, demonstrating real-time translation of Italian speech.
The event also showcased an innovative voice mode that interprets body motion, such as breathing patterns, and generates voice responses in various styles, including robotic and singing voices. The model can respond to comments in a human-like manner, such as reacting to compliments with “Oh stop it, you’re making me blush.”
CEO Sam Altman expressed excitement about the model’s potential, emphasizing its natural interaction and enhanced personalization options. The new model allows users to interrupt and receive real-time responses, reducing the previous 2-3 second delay. It also features emotion detection through facial recognition, exemplified by the AI asking, “Want to share the reason for your good vibes?” when it detected a smiling face.
ChatGPT-4o is also launching with a desktop app that includes voice and vision capabilities. Additionally, the model is integrated into the Artificial Intelligence Application Programming Interface (API), offering developers a more cost-effective and faster option for building their models.
The new ChatGPT-4o supports 50 languages, covering 97% of the world’s population. OpenAI did not address user data protection at the event, though past versions required users to opt-out to prevent their conversations from being used for training. The company uses reinforcement learning from human feedback (RLHF) to address bias in responses. ChatGPT-4o will be available to users within the next few weeks.