Google Announces AI System Gemini After Turmoil at Rival OpenAI

The Alphabet unit said the algorithm wouldn’t be widely available until early next year, citing a need for more extensive safety testing despite spending most of this year on the project.

The announcement of the powerful software, known as Gemini, is the latest attempt by Google to display its AI bona fides after the launch of ChatGPT about a year ago shook up the tech industry and caught the company flat-footed.

Gemini’s release—and public reception—will present a litmus test for Google’s technology following a push to move faster in developing and releasing AI products. It coincides with a period of turmoil at OpenAI that has sent tremors through the tightknit AI community, suggesting the industry’s pecking order is far from settled.

Google wanted outside customers to perform testing on the most advanced version of Gemini before releasing it more widely, said Demis Hassabis, chief executive officer of Google DeepMind.

“We’ve been pushing forward with a lot of focus and intensity," Hassabis said, adding that Gemini likely represented the company’s most ambitious combined science and engineering project to date.

Google said Wednesday it would offer a range of AI programs to customers under the Gemini umbrella. It touted the software’s ability to process various media, from audio to video, an important development as users turn to chatbots for a wider range of needs.

The search company will also use the algorithms to power products such as Bard, its answer to ChatGPT, and mobile-phone features that are capable of running without any network connection.

The most powerful Gemini Ultra version outperformed OpenAI’s technology, GPT-4, on a range of industry benchmarks, Google said. That version is expected to become widely available for software developers early next year following testing with a select group of customers.

Gemini Ultra will also power a new chatbot called Bard Advanced that Google plans to release next year following additional testing. The company didn’t say how it planned to make money from the product.

Google said it would release next week a less sophisticated Gemini Pro offering that would be suitable for a range of tasks. A version of Bard powered by that algorithm, set to be released Wednesday, scored more highly with testers than competitors such as the free version of ChatGPT, Google said.

Google CEO Sundar Pichai announced the Gemini project during a developer conference in May. Executives later used public speaking appearances to tout the system’s capabilities, such as the ability to generate detailed images containing instructions on how to ice a cake.

Google’s co-founder Sergey Brin, an infrequent presence at the company in recent years, began working almost full time again on Gemini with researchers in the newly constructed Charleston East building in Mountain View, Calif., The Wall Street Journal reported.

Pichai tapped Hassabis in April to lead the Google DeepMind research division, a merger of two previously separate teams that sometimes competed for the expensive computing resources required to develop powerful AI algorithms.

Hassabis said Google used data from across the internet to build Gemini, including videos from the company’s YouTube video platform.

Shares in Alphabet have risen around 48% this year, outpacing the broader Nasdaq Composite Index, while lagging behind some tech peers, including Microsoft, OpenAI’s largest backer.

Gemini can process a range of information—including text, computer code, audio, images and video—better than existing AI systems, Google said. The company exclusively used its own specialized AI chips, known as tensor processing units, to develop the systems.

During a demo for the press, Google executives showed a video of Gemini software examining a series of physics homework problems, diagnosing issues in the prewritten solutions and coming up with correct answers.

Google’s announcement came two weeks after OpenAI reinstalled Sam Altman as CEO and appointed new board members, capping a tumultuous period for the startup that prompted some customers to evaluate alternative AI providers.

OpenAI’s GPT-4 algorithm, released in March, is widely considered the most powerful AI system on the market and is used by large companies including Morgan Stanley and Salesforce to build chatbot applications. It powers a subscription version of ChatGPT that costs $20 a month.

Bard’s roughly 220 million monthly worldwide visitors lag well behind ChatGPT’s user base, which is almost eight times as large, according to Similarweb data. OpenAI has also developed a mobile app for ChatGPT, a step Google has yet to take.

Google said it is experimenting with Gemini in a version of its search engine that uses AI to respond with passages of text instead of web links. The product is only available to a few million users, and the company has said it wants to gather more feedback before rolling out the product widely.

Write to Miles Kruppa at miles.kruppa@wsj.com