Google started to roll out its long-awaited Gemini multimodal artificial intelligence model Wednesday morning, touting its ability to natively process and reason across different inputs like text, images, video, and code.
With the new AI model, the search giant is looking to jump ahead in the AI competition against OpenAI, Microsoft, Amazon, and other industry rivals.
Google described Gemini as its largest and most capable model to date, and the first AI model to surpass human experts on the Massive Multitask Language Understanding (MMLU) benchmark.
Sundar Pichai, the Google and Alphabet CEO, said in a blog post that Gemini represents “one of the biggest science and engineering efforts we’ve undertaken as a company,” under the umbrella of the company’s DeepMind AI initiatives.
Google said Gemini has undergone extensive AI safety testing, using tools including a set of “Real Toxicity Prompts” developed by the Seattle-based Allen Institute for AI to evaluate its ability to identify, label, and filter out toxic content.
Gemini will roll out in phases in different products:
- Many Google users will experience Gemini first in the company’s Bard AI chatbot. A version called Gemini Pro will power Bard starting today, Google execs said in a media briefing. A more powerful model called Gemini Ultra will be available early next year in a new version called Bard Advanced.
- Google Cloud will make Gemini Ultra available in an early access program for developers, rolling out more broadly in early 2024. Gemini Pro will be available starting Dec. 13 in Google Cloud’s Vertex AI and AI Studio.
- A version called Gemini Nano for on-device applications will be available on Google Pixel phones, starting with Pixel 8 Pro. Google says it will power a new Summarize feature in the Recorder app and Smart Reply in Gboard.
The Information reported Dec. 2 that Google cancelled a series of in-person launch events that were planned for Gemini “after the company found the AI didn’t reliably handle some non-English queries.”
