https://www.bbc.com/news/videos/cll445rgjq0o
OpenAI has unveiled the latest version of the technology that powers its ChatGPT chatbot. GPT-4o will be available to all users of the previous version, including those who do not pay a subscription to use the service. According to the developers, the new version works faster than the previous one and behaves more relaxed - it can allow itself to chat a little, or even flirt.
GPT-4o can read images and discuss what it sees, translate from one language to another, and detect emotions from facial expressions. The bot, unlike previous versions, can also remember previous conversations. He can be interrupted mid-sentence, and he has an easier conversational rhythm - in the demonstration there was practically no delay between question and answer.
OpenAI said it will roll out new features gradually to ensure they are safe to use. The new model is a step forward in the field of artificial intelligence, says OpenAI chief technical officer Mira Murati. “We're looking at the future of interactions between us and machines,” says Murati. “And we think GPT-4o really changes that paradigm.”
During the demonstration of the voice version of GPT-4o, he not only solved a simple equation written on a piece of paper, but also gave useful recommendations on how to solve it. He analyzed computer code, translating from Italian to English, and interpreted the emotions in a photo of a smiling person.
Following OpenAI, Google's parent company Alphabet showed off its own artificial intelligence efforts. Among other things, she showed the next version of the Gemini chatbot and a new AI assistant for her search engine. Google announced the new developments during its annual developer conference in California.
She focused on the improved Gemini Pro 1.5 model, which is now capable of “digesting” twice as much data. Google also introduced Veo, an artificial intelligence model it claims is the most powerful for creating videos from a simple text command.
The company made its video creation technology available in January, but just a few weeks later Sora from OpenAI, the maker of ChatGPT, came out and turned out to be more advanced. Sora immediately caught the attention of Hollywood, causing both excitement and concern.
Google also announced a "light" version of Gemini called 1.5 Flash, which aims to reduce the cost of AI and increase the speed of results. Like the more powerful version, Flash can handle large amounts of data, but is optimized for chat apps, video, and image captions.
This week, hundreds of millions of Google search users in the United States will begin periodically receiving a summary of their query results, summarized by artificial intelligence technology A.I. Overviews. These "squeezes" are expected to appear at the top of the page only when Google's systems deem them to be the fastest and most effective way to satisfy the user's curiosity.
This most often occurs when the request concerns more complex topics and issues; during a regular search, for example, weather forecasts or goods in stores, the results, as before, will consist of links to websites. Google expects A.I. Overviews will become part of the search results for approximately 1 billion people.
A.I. function Overviews could spell trouble for the company because it could cut into Google's $175 billion in advertising revenue last year alone. The point is that if A.I. Overviews will work too well and will give users everything they need at the top of the page, reducing the number of clicks on links (though they will still appear in search results). As a result, various sites, from large media outlets to entrepreneurs and startups, may lose part of their audience.
Read also:
Olena Zelenskya Met with the President and First Lady of Serbia, Antony Blinken arrived in Kyiv
Pharmaceutical giant AstraZeneca announces worldwide recall of its Covid-19 vaccine