OpenAI, the company behind the popular chatbot ChatGPT, announced on Monday the upcoming release of a new model named GPT-4o. This model is designed to conduct realistic conversations and interact across both text and visual inputs.
The new features include enhanced audio capabilities, allowing users to speak to the robot and receive immediate responses without delays. Users can also interrupt ChatGPT while it is speaking, a functionality that developers have previously struggled to perfect, according to OpenAI during a live-streamed event about the new model.
"It feels like AI from the movies. It has never before felt natural for me to talk to a computer: it does now," wrote Sam Altman, CEO of OpenAI, in a blog post.
Facing stiff competition in the market and under pressure to expand its user base, OpenAI, which counts Microsoft as a major investor, continues to innovate.
Reuters reports that GPT-4o is the latest version of the technology used in ChatGPT, essentially acting as the engine behind the chatbot.
At Monday’s live-streamed event, OpenAI developers demonstrated the chatbot’s new capabilities. For instance, using the model’s visual skills and voice, ChatGPT helped one of the company's employees solve a mathematical equation written on a piece of paper.
In another demonstration, the model was used to translate from one language to another in real-time. When one developer expressed his delight at being able to showcase how "useful and fantastic you (ChatGPT) are," the model humorously responded, "Stop! You're making me blush!"
Following the event, Sam Altman posted a brief message on the social media platform X, stating "Her." According to Reuters, this is a reference to the 2013 movie Her, in which a man falls in love with his AI assistant, voiced by actress Scarlett Johansson.
Mira Murati, OpenAI’s Chief Technology Officer, stated on Monday that the new model will be free because it is cheaper to operate than previous versions.
It will be available in the coming weeks, though specific regions where it will be available were not specified.
