Google launches Gemini, its largest and most capable AI model, to power its products and cloud services

Google announced on Wednesday the launch of Gemini, its largest and most capable artificial intelligence model, which it claims can outperform human experts on a wide range of tasks.

Gemini is a large language model that can understand, operate across and combine different types of information, including text, code, audio, image and video.

Gemini will be available in three different sizes: Gemini Ultra, its largest and most advanced category; Gemini Pro, which scales across a wide range of tasks; and Gemini Nano, which is optimized for specific tasks and mobile devices.

Google plans to license Gemini to customers through Google Cloud for them to use in their own applications, as well as to power its own products, such as its Bard chatbot and Search Generative Experience.

According to Google, Gemini Ultra is the first model to surpass human experts on MMLU (massive multitask language understanding), a benchmark that tests both world knowledge and problem-solving abilities on 57 subjects, such as math, physics, history, law, medicine and ethics. Gemini Ultra can supposedly understand nuance and reasoning in complex subjects.

“Gemini is the result of large-scale collaborative efforts by teams across Google, including our colleagues at Google Research,” wrote CEO Sundar Pichai in a blog post.

“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.”

Starting today, Google’s chatbot Bard will use Gemini Pro to help with advanced reasoning, planning, understanding and other capabilities. Early next year, it will launch “Bard Advanced,” which will use Gemini Ultra, executives said on a call with reporters Tuesday.

It represents the biggest update to Bard, its ChatGPT-like chatbot, since its launch eight months ago.

The launch of Gemini comes amid growing competition in the field of large language models, especially from OpenAI, the research organization co-founded by Elon Musk.

OpenAI launched GPT-3.5, a 175-billion-parameter model, last year, and followed it with GPT-4, a 1-trillion-parameter model, in March this year. Executives said Tuesday that Gemini Pro outperformed GPT-3.5 but did not comment on how it compared with GPT-4.

Google also faced some challenges in developing and testing Gemini, which reportedly delayed its launch.

Eli Collins, vice president of product at Google DeepMind, said that Gemini is the most highly tested AI model that the company has built and that it has “the most comprehensive safety evaluations” of any Google model.

Collins also said that Gemini is more efficient and cheaper to serve than previous models, despite being larger and more capable.

Google said it will release a technical white paper with more details of Gemini on Wednesday, but it will not disclose the number of parameters of the model.

Parameters are a measure of the size and complexity of an AI model, and larger models tend to require more data and compute resources to train and run.

Google hopes that Gemini will enable new possibilities and applications for its customers and users, as well as advance the state of the art in artificial intelligence.

“We believe Gemini will be a powerful tool for researchers, developers, and anyone who wants to explore the frontiers of natural language understanding,” Pichai wrote.

According to Time: DeepMind was founded by Demis Hassabis, Shane Legg, and Mustafa Suleyman in 2010. Google acquired the AI lab for $400 million in 2014, and in April 2023 DeepMind merged with Google’s elite AI research team, Google Brain, to form Google DeepMind, which Hassabis leads.

In 2023, DeepMind merged with Google Brain, another AI research team, to form Google DeepMind, led by one of the founders, Demis Hassabis.

DeepMind’s founders have tried to gain more autonomy from Google and even proposed a new legal structure to prevent a single corporation from controlling powerful AI, but their efforts have been unsuccessful.

Google and Google DeepMind have achieved many AI milestones, such as mastering Go, creating transformer models, and solving protein folding.

However, Google DeepMind has not released any chatbot products, unlike its rivals OpenAI and Anthropic, due to concerns about factual accuracy and quality.

Reference: CNBC

Moody’s and Google Cloud team up to develop AI tools for finance