Google unveils its largest and most powerful AI model, Gemini
Google has unveiled its most advanced artificial intelligence model, Gemini, capable of processing various types of information and running on any device.
This was reported in the Google blog.
The company noted that the Gemini is the most powerful and versatile model they have ever created. Gemini is multimodal, meaning that it can summarize and perfectly understand, operate on, and combine different types of information: text, code, audio, images, and video.
Gemini is optimized for three different variants – Ultra, Pro, and Nano – so it can run anywhere, from data centers to mobile devices. In particular, it is integrated into the latest Google Pixel phones. Pixel 8 Pro is the first smartphone designed to run Gemini Nano.
Gemini Pro has already been tested by a number of industry standard tests. Gemini Pro outperformed GPT-3.5 in six out of eight of them, including better performance in MMLU (Massive Multitask Language Understanding), one of the standards for measuring large artificial intelligence models, and in GSM8K, which measures the mathematical thinking of elementary school students.
Google said that Gemini will be integrated into Bard, Google’s AI-powered chatbot in English available in more than 170 countries. The model will be launched at Bard in two stages: first, the chatbot will use a specially customized version of Gemini Pro in English for more advanced reasoning, planning, understanding, etc. And early next year, the company plans to introduce Bard Advanced, which will provide access to more advanced models and features, starting with the Gemini Ultra.
For the time being, Gemini will not be available in Europe and the UK. Google noted that this is due to regulatory obstacles.
“We’re working on this and working with local regulators to make sure we engage with them before launching a product in any particular area,” Google said.
