Google Delays Release of Gemini AI That Aims to Compete With OpenAI

Google’s company-defining effort to catch up to ChatGPT creator OpenAI is turning out to be harder than expected.

Google representatives earlier this year told some cloud customers and business partners they would get access to the company’s new conversational AI, a large language model known as Gemini, by November. But the company recently told them not to expect it until the first quarter of next year, according to two people with direct knowledge. The delay comes at a bad time for Google, whose cloud sales growth has slowed while that of its bigger rival, Microsoft, has accelerated. Part of Microsoft’s success has come from selling OpenAI’s technology to its customers.

The Takeaway
• Google Cloud plans to let enterprise customers use Gemini early next year
• Researchers are scrambling to make sure the model can surpass GPT-4
• Google co-founder Sergey Brin has taken on an expanded role on Gemini team

Google’s difficulties in launching Gemini to consumers and enterprise customers have hurt it in more ways than one. Since OpenAI launched ChatGPT a year ago, the startup has ascended to the tech industry’s peak while turning the chatbot into a global synonym for AI among consumers. And OpenAI did so with technology that Google invented and with the aid of employees who previously worked at Google.

ChatGPT also is the backbone of one of the fastest-growing consumer and enterprise software businesses in recent memory. Companies use the technology for tasks like automating software coding, summarizing long reports, generating marketing campaigns and building specialized apps that make use of its predictive capabilities. (On Tuesday, OpenAI CEO Sam Altman said his company had run out of server capacity following a surge of sign-ups for the paid version of ChatGPT.)

Meanwhile, Google’s own ChatGPT rival, Bard, has failed to take off. That’s a problem for Google because ChatGPT’s millions of customers generate valuable data that helps OpenAI track and improve the quality of its products.

“We’re not commenting on rumors or speculation,” Google spokesperson Kathryn Watson said. On Thursday, Google CEO Sundar Pichai, who in May confirmed The Information’s earlier reports about Gemini, said at a public event that the company is “focused on getting Gemini 1.0 out as soon as possible, make sure it’s competitive, state of the art, and we’ll build from there on.”

Google’s delay of the large Gemini model for cloud customers implies that the company wants to buttress its consumer offerings with the new tech before giving outside software developers access to it, one person said.

On a call with investors on Oct. 24, Pichai didn’t explicitly discuss Gemini’s launch but said, “We are just really laying the foundation of what I think of as the next-generation series of models we’ll be launching all throughout 2024.”

High Hopes

Google has high hopes for Gemini that go beyond boosting enterprise software sales. The company wants it to power new tools for creators on YouTube, such as the ability to generate custom backgrounds for videos, and to improve the capabilities of Bard as well as Google Assistant, Google’s Siri-like voice assistant for phones and other devices.

Google has developed several versions of Gemini to handle different tasks, depending on their complexity. Outside developers have already tested smaller versions of the model, measured in terms of the number of parameters, or calculations they do. But the company is still finalizing the primary, biggest version of Gemini, said a person who has been involved in the effort. (A big LLM typically is made up of smaller models that work together.)

A key challenge for the Gemini team is making sure the primary model is as good as or better than OpenAI’s most advanced LLM, GPT-4. It’s not clear whether Google has met that standard, this person said.

The Gemini developers have had help from co-founder Sergey Brin, who returned to the company to work on it and now spends four to five days a week with the model’s developers in the company’s Mountain View, Calif., headquarters.

Comparing the quality of AI models from different companies isn’t simple, especially if one of them hasn’t launched publicly. That’s because Google isn’t getting the same amount of user feedback about Gemini’s quality that OpenAI gets from customers of its GPT models. OpenAI, meanwhile, has had its own challenges in improving GPT-4 meaningfully, and it quietly canceled a major new model it was developing earlier this year, called Arrakis.

Google has taken a cautious approach to using Gemini in Bard, which a less capable LLM has been powering since Bard launched in March. Bard made a factual error in a closely watched demonstration at that time, and that mistake still haunts some employees involved with it, said a Google employee.

AI for Advertising

Another key application for Gemini involves advertising, Google’s main moneymaker. Google executives have discussed using the new model to power tools for advertisers that can automatically generate ad campaigns, including custom text and visuals, a person familiar with the matter said. That will initially include images for static display ads, but it could eventually encompass audio and video ads as well. (Other companies, including Amazon and Meta Platforms, also have launched AI-powered ad tools.)

Gemini also has a longer memory of a user’s interactions, compared to earlier Google LLMs such as Palm 2, which currently powers Bard and generative AI results in Google Search. The longer memory could allow advertisers to compare the performance of campaigns over time, the person familiar with the matter said. For instance, an advertiser could use the model to create new variants of the best-performing ad copy over the last month, this person said.

Google has had all year to work on Gemini. But the endeavor required calling a truce between two competing AI teams, Google Brain and DeepMind, which was a stand-alone unit within Google’s parent company Alphabet. The teams eventually merged. (See the Gemini team here.)

The Gemini developers have had help from co-founder Sergey Brin, who returned to the company to work on it and now spends four to five days a week with the model’s developers in the company’s Mountain View, Calif., headquarters, a person with knowledge of the situation said. He isn’t formally a decision-maker, this person said, but in recent weeks he has offered criticism and feedback on Gemini and has helped different teams coordinate their work. Brin has a controlling stake in Google’s parent company, Alphabet, together with co-founder Larry Page.

Google and OpenAI have been battling fiercely over AI talent, given the dearth of people with expertise in developing LLMs. OpenAI has been offering annual pay packages that could be worth $10 million or higher for certain researchers at Google, The Information reported. One factor for the AI researchers weighing such offers is computing capacity. Google holds an edge over OpenAI in terms of the raw amount of specialized servers employees can use to develop new AI models. OpenAI CEO Sam Altman told colleagues earlier this year that Google would hold onto that kind of advantage into next year, but that OpenAI hoped to catch up with the aid of Microsoft, whose servers power OpenAI’s technology.

Jon Victor is a reporter at The Information covering Alphabet. He can be reached at jon.victor@theinformation.com or on Twitter at @jon_victor_.