OpenAI has just rolled out GPT-4o, its most advanced AI model yet, and it's designed to make your digital interactions feel almost human. During the announcement, OpenAI's team gave a live demo of the new GPT. Going by that interaction, it is safe to say that the new Voice Mode will get exponentially better after the update, making the chatbot sound almost like a human being. The team constantly kept interrupting the ChatGPT while it was trying to respond and it kept modifying its response accordingly.
So, to understand what's going on under the hood, let's dive into the details of this new update and explore why it could be OpenAI's most intriguing and impactful update yet.
Faster and more fluent One of the standout features of GPT-4o is its speed. Not only does it match the prowess of its predecessor, GPT-4 Turbo, in handling English text and coding tasks, but it significantly betters the handling of non-English languages. This means a smoother experience for a global user base.
Blending sight, sound, and text GPT-4o isn’t just about text. It integrates audio and visual inputs and outputs too. Imagine asking your computer a question out loud and it recognising not just your words, but the tone and context, or showing it a picture and getting an explanation in seconds. GPT-4o can respond to spoken queries in as little as 232 milliseconds—comparable to a human response in a conversation.
More seamless processing Earlier versions of Voice Mode in ChatGPT involved a somewhat clunky process where different models handled different tasks: one model would transcribe speech to text, another would process the text, and yet another would turn the text back into speech. GPT-4o simplifies all this with a single model handling text, vision, and audio from start to finish. This not only cuts down on response time but also improves the quality of the interaction. The model can now detect nuances like tone, recognise multiple speakers, and even incorporate sounds like laughter or singing into its responses.
When will you get to play with the new GPT-4o? Starting today, GPT-4o’s capabilities are being integrated into ChatGPT, initially in text and image formats, with plans to roll out audio and video capabilities to select partners soon. It's available in the free tier and to Plus users, who will enjoy up to five times higher message limits.
Safety features OpenAI claims it has implemented new techniques to ensure the model’s outputs remain reliable and safe across all new modalities. This includes refined training data and built-in safeguards specifically designed for voice interactions. OpenAI is also inviting feedback to refine and improve GPT-4o.