Artificial intelligence developer OpenAI has announced its new multimedia language model called GPT-4o. The model will be presented in the coming weeks at the ChatGPT exhibition, which in November 2022 opened a new era in the field of smart services that generate conversations and texts.
According to the developer, GPT-4o understands 50 languages and communicates more naturally. According to the published explanatory video, he laughs at the user's jokes.
Until now, GPT-4 has been operating under ChatGPT, and the letter “o” is the name of the new generation Omni It covers the word, indicating that the robot, in addition to the previous text mode, can now recognize and generate sound and image. In an app for blind people, for example, it can describe the environment verbally on the spot.
The GPT-4o is faster than the GPT-4 Turbo that was introduced at the end of 2023 – according to the company, it responds in just 320 milliseconds instead of 2-3 seconds.
Mira Moratti, CTO of OpenAI, said at the press presentation that they want to make AI accessible, so free GPT-4o-ta ChatGPT users who were familiar with GPT-3.5 will be able to try it out. The free version is limited in computing capacity, and subscribers will be able to work with five times that limit.
With the announcement, it was also decided that the little-known software that appeared in the LMSys chatbot scene, “I'm-also-a-jo-gpt-2-chatbot-bayok” is none other than GPT-4o and will soon appear at the top of the leaderboard.
According to expert reviews, in addition to new useful abilities, the model's tendency to hallucinations has not changed, that is, its habit of making ridiculous statements out of nothing has remained.
The head of the security team leaves
Shortly after announcing the arrival of the new GPT, Ilya Sutskever, OpenAI's founder and head of research, announced that he was leaving the company. Based on his post on X, Sutskever is ready to step down and will be working on a personal project close to him in the near future, the details of which he will share later.
Sutskever, who was invited by Google's Elon Musk to OpenAI in 2015, was one of the heroes of the palace revolt that broke out at the company half a year ago, which began with the dismissal of CEO Sam Altman and then ended in confusion, as employees were also threatened with dismissal. As a result of the events, disagreements arose between Altman and Helen Toner, Director of Strategy. Sutskever quickly expressed regret for supporting the ouster of Altman, who returned to head the company days later.
Incidentally, the controversy that sparked the scandal was about the social consequences and security problems of artificial intelligence work. Sutskever led the so-called Superalignment team, which had to ensure that the software behaved as expected, and was not toxic, hostile, criminal or harmful in any way.
Altmann described Sutskever as his friend and one of the greatest minds of his generation in his own statement on X. Also leaving with the specialist was his close colleague Jan Laecke.
(Ars Technica, CNBC, CNET, GSMArena)
Follow Index on Facebook too!
