OpenAI announced on Monday that it has introduced an update to its artificial intelligence program, “Chat GPT,” that will make it able to interact with users in audio and video.
Soon, the program made famous by generative artificial intelligence (the ability of software to produce text, images and other content based on a simple request in everyday language), will be able to respond to any request associated with an image in addition to chatting verbally with its users.
For example, they will have the opportunity to take a photo of a monument and “have a conversation with GBT chat” about the history of this landmark, or show it the ingredients in their refrigerator to suggest a recipe, as Open AI indicated in a statement.
The startup listed other uses for the program, including helping children do their homework (by taking a picture of a math problem, for example) or asking children to tell them a bedtime story.
These updates will be introduced to “Chat GPT” during the next two weeks, while subscribers to “Chat GPT Plus”, that is, the paid version of the program, and institutions benefiting from the service will benefit from them.
The company announced that it intends to make these updates to the program, during its presentation in March of “Chat GPT 4”, the latest version of the program.
GPT4 Chat is multimedia, meaning it can process data other than text and code.
The success achieved by ChatGPT in 2022 prompted major technology companies such as Google and Microsoft to race to create programs that work based on generative artificial intelligence.
But the rapid deployment of these programs, which are not yet subject to clear rules, raises great concern, especially since they invent answers to any question asked of them.
OpenAI acknowledged in its statement on Monday that “the models that will be able to process images represent new challenges, because people may use the image captions that the program will provide in precise and risky areas.”
The startup confirmed that it had “tested the model” on topics including extremism and scientific knowledge, expressing its aspiration to use it in daily life and receive comments from users to improve its performance.
On the other hand, the company limited GBT Chat's capabilities to "analyze people" because the interaction "is not always accurate," stressing that "these programs must respect individuals' privacy."
In addition, Spotify announced on Monday a partnership with OpenAI to translate podcasts directly using artificial intelligence.
She indicated in a statement that episodes shown in English on the platform will be available in other languages, “while preserving the distinctive vocal characteristics of the speaker.”
The Swedish company confirmed that the startup's new sound generation technology "replicates the style of the original speaker, allowing the user a voice copy that is very close to the original, and more realistic than traditional dubbing."