Google Unveils Gemini, Its Massive AI Model And Biggest Threat To ChatGPT Yet
Google and Alphabet CEO Sundar Pichai remarked in a Google blog post, “This is what excites me: the chance to make AI helpful for everyone, everywhere in the world.” He continued to add that Google is approaching the advancement of AI “boldly and responsibly.” He noted Gemini is the company’s most capable and general model to date, with Gemini 1.0 being optimized for different sizes: Ultra, Pro, and Nano.
Google explains the three sizes as such: Gemini Ultra is Google’s largest and most capable model for highly complex tasks. Gemini Pro is the company’s best model for scaling across a wide range of tasks. Gemini Nano is the most efficient model for on-device tasks.
Demis Hassabis, CEO and Co-Founder of Google DeepMind, remarked, “Gemini is the result of large-scale collaborative efforts by teams across Google, including our colleagues at Google Research. It was built from the ground up to be multimodal, which means it can generalize and seamlessly understand, operate across and combine different types of information including text, code, audio, image and video.”
The most impressive benchmark noted by Google is perhaps its score of 90% on MMLU (massive multitask language understanding). Google touts Gemini Ultra as being the first AI model to outperform human experts across a combination of “57 subjects such as math, physics, history, law, medicine, and ethics for testing both world knowledge and problem-solving abilities.”
The video above from Google highlights some of the company’s favorite interactions with Gemini. It showcases a wide range of challenges, showing Gemini a series of images, and then asking it to reason about what it sees.
According to Google, Gemini 1.0’s multimodal reasoning capabilities are skilled at uncovering knowledge that can be difficult to discern a wide range of data. The company says that it has the ability “to extract insights from hundreds of thousands of documents through reading, filtering and understanding information” which will help deliver “new breakthroughs at digital speeds in many fields from science to finance.”
Gemini 1.0 can also understand, explain, and generate high-quality code in programs such as Python, Java, C++, and Go. Google says this ability to work across multiple languages and being able to reason about complex information, makes it an ideal candidate for coding around the world.
Gemini Pro is currently available to use through Google products such as Bard. Gemini Nano is available to use with a Google Pixel 8 Pro, with more updates coming to the Pixel portfolio of phones. Gemini Ultra is still undergoing extensive trust and safety checks, but should be made available to developers and enterprise customers early next year.
To read a full breakdown of Gemini and its capabilities, head to Google’s website. Be sure to stay tuned to HotHardware, as we will continue to cover Gemini as it continues to evolve.
