Rumor: OpenAI is working on an AI voice assistant that can recognize images – IT Pro – News

OpenAI is said to be working on a “multimodal digital AI assistant.” Users can have conversations with it and the assistant can recognize objects in photos. The company may announce the voting form as early as Monday.

According to sources of information This new multimodal model can understand audio “faster and more accurately” than OpenAI’s current text-to-speech model. The site writes that the AI ​​product can, among other things, better understand the tone of speakers, so that it can detect, among other things, whether they are being sarcastic. This should be useful for business applications, such as automated customer service, for example, according to the information.

In addition, the tool should be able to recognize objects that users photograph, just as is already possible with Google Gemini. According to the sources, “The model could help students with their mathematics homework, translate signals in the real world, or solve car problems.” Info wrote that the model can answer “some types of questions” better than GPT-4 Turbo, though it doesn’t explain more.

According to the sources, this model will not be announced until Monday at the earliest. On that day, OpenAI will be held at 7 p.m It is a live broadcast. App developer Ananay Aurora Discovered references In the ChatGPT code it is indicated that there will be a feature that will allow users to make phone calls within the tool. Aurora expects this feature to also be announced on Monday.

Reuters sources said earlier this week that the artificial intelligence company would announce its search engine that day. Friday denied OpenAI CEO those rumors. He also said that GPT-5 will not be revealed during the event. The latest model will be released publicly later this year, The Information wrote.