(San Francisco) Less than a year after the phenomenal launch of the generative artificial intelligence (AI) interface ChatGPT, the start-up OpenAI on Monday presented its latest innovations to make its technology more efficient, personalized, easier to use and also more expensive for developers .
Posted at 3:11 p.m
“On November 30th, we discreetly put a first version of ChatGPT online for research purposes. And that went pretty well,” joked Sam Altman, the company’s boss.
“We now have around 100 million active users every week,” he added on stage in San Francisco during a conference broadcast live online.
The unprecedented success of ChatGPT and the concerns raised by generative AI have brought the young leader (38 years old) to the forefront in 2023, from parliamentary hearings to interviews with heads of state.
On Monday, he reiterated his confidence in the ability of AI in the future to give every human being emancipatory abilities “on a scale we have never seen before.”
“We will be able to do more, create more and have more,” he assured. “With (artificial) intelligence integrated everywhere, we will all have superpowers when needed. »
Because OpenAI introduced “GPT” on Monday: Users can now create personalized dialogs without programming. “For example, you can use GPTs to learn the rules of a board game or teach your children math. »
These conversational robots are based on OpenAI language models (the core technology of generative AI) and on instructions and documents added by the creator of the dialogue (like the rules of a game).
Widely viewed as a revolution comparable to the advent of the Internet, generative AI was popularized by ChatGPT. It makes it possible to produce texts, images and sounds in everyday language upon simple request.
For many observers, this will make it possible to create personalized AI agents that support people in their personal and professional lives.
In addition to GPT, OpenAI’s announcements were primarily aimed at the more than 2 million developers who use its technologies to build applications.
Sam Altman introduced “GPT-4 Turbo,” a new model that can take more context into queries and have Internet access by April 2023 (instead of 2021).
The application programming interface (API) gains multimedia capabilities (computer vision, speech, etc.). And the prices for using the models are falling.
“We hope you come back next year because what we brought to market today will pale in comparison to what we create for you,” promised Sam Altman.