Finance Associated Press, December 7 (edited by Niu Zhanlin).On Wednesday local time, the American technology giant Google announced the launch of what it considers to be the largest and most powerful artificial intelligence model Gemini, which is capable of processing information in different content forms such as **, audio and text.
Google says its highly anticipated AI model, called Gemini, is capable of making more complex inferences and understanding more nuanced information than previous technologies. By reading, filtering, and understanding information, it can extract gist points from hundreds of thousands of documents and will help achieve new breakthroughs in many fields, from science to finance.
"This new model represents one of the biggest scientific and engineering efforts we've made as a tech company, and it's also a multimodal foundational model that generalizes and understands different types of information, including text, audio, images, and more," Google CEO Sundar Pichai wrote in a blog post.
Since OpenAI launched ChatGPT a year ago, Google has struggled to develop AI software that can compete with the company. Google claims that it has added technology from some Gemini models to its AI assistant Bard, and says it plans to fully integrate the state-of-the-art Gemini model into Bard early next year.
Google executives believe that the Gemini Pro outperforms GPT-35, but dodging the question of how it compares to GPT-4. And in March of this year, OpenAI launched GPT-4.
The tech company said it will release three versions of Gemini, namely Gemini Ultra, Gemini Pro and Gemini Nano. Each version has different information processing capabilities, with the most powerful Gemini Ultra version designed to run in the data center, and the weakest Gemini Nano version that will run efficiently on mobile devices.
Starting December 13, developers and enterprise customers will be able to access Gemini Pro through the Gemini API in Google AI Studio or Google Cloud Vertex AI. Android developers can also use Gemini Nano for software development.
Eli Collins, VP of product at DeepMind, claims that Gemini is the most powerful AI model that Google's DeepMind AI unit has helped create, but that it provides users with a "significantly" cheaper service compared to the company's previous large models.
Collins adds: "As a result, Gemini is not only more powerful, but also much more efficient. The latest models still require a lot of computing power to train, and Google is moving this process forward at a rapid pace. ”
Google has also released its most powerful AI chip, Cloud TPU V5P, which is an improvement on the previous version. According to Google, TPU V5P has a two-fold improvement in floating-point performance compared to TPU V4, and it trains large language models 2.2 times faster than TPU V48 times.