Google has officially announced the launch of its multimedia AI system, Gemini, which competes with OpenAI's GPT-4 model.
According to Google official blog, the new system is designed to be multi-media, as it is trained on different types of data and then improved using additional data to further improve its effectiveness, which helps Gemini understand all types of data. Write and think seamlessly and combine input types. Different types of information such as text, code, audio, images and videos.
Google announced that the Gemini system will be offered in three different versions to meet the needs of different users. The “Ultra” version is the largest and most efficient of these versions, and is designed for complex tasks, followed by the “Pro” version that can be used on a variety of devices, and then comes the “Nano” version for tasks that can be performed on a single device.
In terms of performance, Google claims that the Ultra version of Gemini outperforms 30 out of 32 academic benchmarks for evaluating large language models (LLM), and also claims that Gemini outperforms existing multimedia models, most notably OpenAI's GPT-4 model in Almost all fields. Models. Evaluation standard.
Google said it is working to add more security measures to reduce potential risks while improving new system features.
Google has begun introducing artificial intelligence systems into several products, including its chatbot “Bard,” which will be built on a modified version of the Pro version. The Pixel 8 Pro phones are also getting a Nano model to perform certain tasks, such as collecting information from Google. Gemini will add to the new generative research experience.
Google said it will delay the launch of the Ultra version due to the need for “more in-depth security reviews” but will allow early access for some customers early next year.