Andreas Braun, CIO of Microsoft Germany, announced that the company plans to launch next week a new generation of the GPT-4 artificial intelligence model, which is an improved version of the GPT-3.5 model that originated on the basis of the popular chatbot. chat
According to German website Heise, the announcement was made at the company's artificial intelligence technology conference in Germany.
"We will launch GPT-4 next week and we will have several models with different capabilities just like video," said the site (Brown), adding that the new version has become more advanced in supporting all languages and indicates that users can, for example, ask questions in German and receive answers in Italian. He added that the variety of models would make the technology "more complete".
The GPT model used in artificial intelligence is said to have been primarily developed by OpenAI, but Microsoft has invested $10 billion in the startup in return for the right to modify the company's model for use in its products, which the company announced last month. Bing ChatGPT engine. The company called its model Prometheus and said it was superior to traditional ChatGPT services.
OpenAI previously announced that it is developing a GPT-4 model, which will be more powerful and faster than GPT-3.5, but it did not provide many technical details about it or its release date, but previous reports stated that it is expected to be announced sometime this year.
OpenAI currently offers several different AI models, including DALL-E 2 that allows users to create images with text, Whisper, an engine that converts speech to text, and the ability to type text and answer questions from GPT.
So far, ChatGPT cannot be used to create images or videos, but (Brown) confirmed that its next version will be "multimedia", which means that it can be used to create different types of data, including videos.
Last year, Google announced the development of Imagen technology, which can create video clips from text sentences, but the service is still in development and is not yet available to all users.