The research departmentDeepMindofGooglejust launched two days agoGemini. It has been developed in three variants, which are Ultra, Pro, and Nano. The Pro version also haswas integrated into Bardin more than 170 countries, but not yet in Europe.Alphabetfurther indicates that this should happen quickly.
Never seen before
ConcerningGemini Ultra, which is the most elaborate version of the model, it exploded the scores of all the performance tests. This by obtaining results superior toGPT-4on almost all tests. This had never happened before. From a purely functional point of view, its performance allowed it, for example, to analyze and synthesize200,000 scientific research articlesin just one hour.
For comparison,Elon MuskpresentedGrok-1early last month, andthe results were very much lowerto those presentedGooglea few days ago.Geminiis also a model“multimodal”, that is to say, it is designed to understand any type of query, regardless of its format. Whether it is text, image, video, audio or code, he will know how to process it and provide an adequate response. In order to prove these allegations,Googlemade it big by posting more than 17 videos on his channelYouTubewhich demonstrate the capabilities ofGemini.
Future ally of the visually impaired?
In one of them, we can see the AI interacting with a human being who submits various visual requests to it in real time. The AI describes what is happening on the screen without any problem. Which can be ideal for helping peoplepoorly or blindly. Here is the video, it is in English but subtitles are available in French.
Everything you need to know in 90 seconds
Given the significant number of new features deployed,Googleoffers a 90-second summary video to get an overview of the scope ofGemini. Other videos showing more specific uses ofGeminiare available on the channelYouTubeofGoogle.
i-nfo.fr - Official iPhon.fr app
By : Keleops AG
