OpenAI has a new ChatGPT model with an improved voice assistant and free users will have access to custom chatbots for the first time.
OpenAI has announced the launch of a new model that will bring the same experience to all users, including users of the free version of ChatGPT, as well as a great voice interface that could rival Amazon's Alexa.
In the so-called “Spring Update,” OpenAI introduced the new ChatGPT-4o model in its first live event on Monday.
Mira Moratti, CTO at OpenAI, told the audience that the model is much faster than previous ChatGPT-4 and that it offers improvements in text, video and audio.
“It's a big step forward in terms of ease of use,” he said, explaining how he could instantly translate his speech into Italian.
What else can ChatGPT-4o do?
The long-awaited event also revealed a voice mode that can read body movement, the intensity of the user's breathing, and can generate sound with different emotional styles on request, such as the voice of a robot or a singer.
He also responds to comments in a humane way, for example, when he was praised for being “helpful and cool”, he replied: “Oh, stop that, you're making me blush.”
“Talking to a computer never felt natural to me; now it does. As we add (optional) personalization, access to information, the ability to take action on your behalf, and more, I can see an exciting future where we can take advantage of it,” said Sam Altman. , OpenAI CEO and co-founder, said in a blog post: “Computers to do so much more than ever before.”
The company stated that, unlike previous versions, users can interrupt the model and it can respond in real time, reducing the time lag to 2-3 seconds.
ChatGPT is now also able to detect emotions when looking at a face through the camera. During the demonstration, they showed a smiley face and the AI asked them: “Do you want to share why you are in a good mood?”
ChatGPT is also launching a desktop application with audio and vision capabilities.
Available to everyone and faster
Another big update is that the model is now available on the AI Application Programming Interface (API), which means developers can start building their model at 50 percent lower cost and twice as fast.
ChatGPT-4o is also available in 50 languages, covering 97% of the world's population.
Data protection and ethics?
OpenAI did not mention during the event whether it would protect user data. In previous versions, ChatGPT could use your conversations for training unless you chose not to.
Generative AI has been criticized for AI biases and hallucinations (incorrect and misleading responses). The company did not say how or whether it would improve this aspect in the new model.
Currently, OpenAI uses a method called Reinforcement Learning from Human Feedback (RLHF), which occurs when humans analyze biases in chatbot responses.
When will it be available?
OpenAI stated that ChatGPT-4o will be available to ChatGPT users in the coming weeks.
“Coffee trailblazer. Social media ninja. Unapologetic web guru. Friendly music fan. Alcohol fanatic.”