Improvements are being rolled out. OpenAI indicated on Monday, September 25, that it had equipped its artificial intelligence (AI) program ChatGPT with speech and vision to make it “more intuitive”.
The interface that made generative AI popular (capable of producing text, images and other content on simple request in everyday language) will soon be able to process requests containing images and also chat orally with its users.
They will be able, for example, to take a photo of a monument and “have a conversation with ChatGPT” about the history of the building, or even show the software what is in their fridge so that it can suggest a recipe. , suggests OpenAI in a press release.
Helping your children with their homework
Other possible use cases according to the start-up: helping your children do their homework (by taking a photo of a math problem for example) or even asking the chatbot to tell them a story before going to sleep.
These new tools will be deployed in the next two weeks for subscribers to ChatGPT Plus, the paid version of the chatbot, or customer organizations of the service.
The company announced the upcoming addition of such features last March, at the time of the presentation of GPT-4, the latest version of its language model, the technology which underlies chatGPT. GPT-4 is multimedia, in the sense that it can process data other than text or computer code.
In the race for generative AI
The success of ChatGPT since the end of 2022 has led to a major race for generative AI between technology giants, with Google and Microsoft in the lead. But the rapid deployment of these still very poorly regulated programs also raises a lot of concern, especially since they have a tendency to “hallucinate”, that is to say, to invent answers from scratch.
“Models with vision present new challenges, from hallucinations to having people rely on the program’s interpretation of images in high-stakes domains,” OpenAI acknowledged in its statement Monday.
The start-up claims to have “tested the model” on subjects such as extremism and scientific knowledge and is counting on real-life uses and user feedback to improve.
It further limited ChatGPT’s abilities to “analyze people” because the interface “is not always accurate and these systems must respect the privacy of individuals.”
A partnership with Spotify
The streaming platform Spotify also announced on Monday a partnership with OpenAI to translate podcasts directly with AI.
Broadcasts recorded in English will now be available in other languages ”while retaining the speaker’s distinctive vocal characteristics,” the service said in a statement.
The Swedish company assures that OpenAI’s new voice generation technology “reproduces the style of the original speaker, allowing for a more authentic, more personal and more natural listening experience than traditional dubbing.”