'Hey ChatGPT' - OpenAI gives its AI a voice for spoken conversations

DPA
·2 min read
A helpful AI that is barely indistinguishable from a human being has so far been the stuff of science fiction, but now that ChatGPT has a voice, technology appears to have taken a leap. Silas Stein/dpa
A helpful AI that is barely indistinguishable from a human being has so far been the stuff of science fiction, but now that ChatGPT has a voice, technology appears to have taken a leap. Silas Stein/dpa

The world's most famous chatbot now has a voice and can use it to have real-time conversations with humans, developer OpenAI demonstrated on Monday, unveiling new features of ChatGPT's AI model GPT-4.

In an update that makes ChatGPT a serious competitor to the voice assistants of tech heavyweights like Google, Amazon and Apple, the AI chatbot can now interact with users in a lifelike human voice.

Whereas ChatGPT was previously largely used in a text chat, the app can now also respond verbally to what it hears, as well as what it sees through a smartphone camera.

In a live demonstration, OpenAI showed how the new model can help solve a maths equation. Instead of just telling the user the answer to the problem, the software gives advice on how to solve it.

One OpenAI employee says he will write an algebra equation on a piece of paper and point the smartphone camera at it to get help solving it.

ChatGPT initially makes a mistake and says it can see the equation before the employee has even started writing - a small slip-up for OpenAI, but a moment that makes the demo appear all the more genuine.

"What do you think we should do with that +1?" ChatGPT asks in a friendly teacher-like tone after giving a brief explanation of the equation. In another demo, ChatGPT invents a bedtime story and reads it aloud.

Another employee demonstrates that you can interrupt ChatGPT just by talking, unlike with rivals Google and Amazon, which tend to keep on talking.

You can also tell it to add more drama to its voice or speak like a robot. Throughout the demonstration, ChatGPT speaks fluently and vividly like a human. A live demonstration with ChatGPT translating a few sentences between Italian and English also worked smoothly.

All interactions can be started with the wake-up phrase "Hey, ChatGPT". The voice assistant update is now available in the ChatGPT app.

OpenAI's technology executive Mira Murati emphasised that the new GPT-4 model will also be available for free users. Until now, the AI start-up has tended to include extended functions in the paid subscription version.

The timing of the presentation also is interesting, as Google is holding its annual developer conference on Tuesday, at which announcements on new functions with artificial intelligence are also expected. A few months ago, Google showed how its Gemini AI software can process spoken and visual information.