Google has announced the launch of Gemini, a new multimodal AI model family aimed at rivaling OpenAI’s GPT-4. The largest version of Gemini is said to exceed current state-of-the-art results on 30 of the 32 widely used academic benchmarks in large language model research. A specially tuned English version of the mid-level Gemini model is already available in over 170 countries as part of the Google Bard chatbot, although not in the EU or the UK due to potential regulation issues.
Gemini can handle multiple types of input, making it multimodal, and can process text, code, images, and audio. It aims to deliver accurate problem-solving, advice, and answers across various fields, from mundane to scientific. Google plans to integrate this technology tightly into its products, highlighting Gemini as the first realization of the vision formed earlier in the year when Google DeepMind was formed, according to CEO Sundar Pichai. The Gemini model comes in three sizes: Ultra, Pro, and Nano, each tailored for different levels of complexity and tasks.
Google’s new Gemini AI model family, which aims to rival OpenAI’s GPT-4, can handle multiple types of input, including text, code, images, and audio. The technology aims to deliver accurate solutions, advice, and answers across various fields. Google plans to integrate this technology into its products, with Sundar Pichai calling it one of the biggest science and engineering efforts the company has undertaken. Gemini comes in three sizes, with only the mid-level model currently available for public use. Lastly, Google claims that Gemini is more scalable and efficient than its previous AI models when run on its custom Tensor Processing Units (TPU), and excels in the area of coding.