Sergey Brin says Google “definitely messed up” Gemini launch

Google Co-founder Sergey Brin, in a rare public appearance over the weekend, told a group of AI enthusiasts that he was coming out of retirement “because the path of AI is so exciting.”

Brin, 50, spoke to entrepreneurs on Saturday at AGI House in Hillsboro, California, just south of San Francisco, where developers and founders were testing Google's Gemini model. AGI stands for Artificial General Intelligence and refers to a form of artificial intelligence that can complete tasks at the same level as humans or a step above.

While answering audience questions, Brin discussed the impact of AI on search and how Google can maintain its leadership position in its core market as AI continues to grow. He also commented on last month's flawed launch of Google's image generator, which the company pulled after users discovered inaccurate information and questionable responses.

“We definitely messed up the image generation process,” Breen said Saturday. “I think it was mostly due to the lack of comprehensive testing. It certainly, for good reasons, upset a lot of people.”

Google said last week that it plans to relaunch the photo creation feature soon.

Brin co-founded Google with Larry Page in 1998, but resigned as chairman of Alphabet in 2019. Brin remains a board member and a major shareholder, owning a stake in the company worth about $100 billion. He has returned to work at the company as part of efforts to help strengthen Google's position in the highly competitive artificial intelligence market.

In some cases on Saturday, Breen said he was providing “personal” answers, rather than representing the company.

“Seeing what these models can do year after year is amazing,” he said at the event, a recording of which was watched by CNBC.

Regarding recent challenges with Gemini that resulted in flawed image results, Brin said the company isn't entirely sure why the responses skewed left, in a political sense.

“We don't fully understand why it leans leftward in many cases,” he added, but “that's not our intention.” Breen added that the company recently made accuracy improvements of up to 80% in some internal tests.

Brin's comments mark the first time a company executive has spoken about the Gemini issue in a face-to-face interview. The company previously sent prepared statements from Prabhakar Raghavan, Google's head of search, and CEO Sundar Pichai in response to the controversial rollout.

Here's what Raghavan said in a blog post on February 23:

“So what went wrong? In short, two things. First, our adjustment to ensure that Gemini showed a range of people failed to account for cases that clearly did not show a range. Second, over time, the model became more cautious than we intended and rejected Answering some prompts completely – Misinterpreting some very soothing prompts as sensitive. These two things led to the model being overcompensated in some cases, and overly conservative in others, resulting in awkward and wrong images.

Google declined to comment on this story. Brin did not immediately respond to a request for comment.

Brin said that Google is not alone in its struggle to achieve accurate results using artificial intelligence. He cited OpenAI's ChatGPT and Elon Musk's Grok services as AI tools that “say some very strange things out there that definitely seem on the far left, for example.”

Hallucinations, or incorrect responses to a user prompt, are still “a big problem right now. There's no doubt about that,” he said.

“We've been making them hallucinate less and less over time, but I would certainly be excited to see a breakthrough close to zero,” Breen said. “But you can't just rely on breakouts, so I think we'll continue to do the additional things we're doing to bring it down over time.”

When an audience member asked him if he wanted to build artificial general intelligence, Brin answered in the affirmative, citing AI's ability to help with “inference.”

Brin was also asked how to disable online advertising since ad revenue is key to Google's business. The company has reported advertising growth has slowed in the past few years.

“I of all people am not terribly concerned about business model shifts,” Breen said. “I think it's great that we've now been able for 25 years, or whatever, to offer world-class information search for free to everyone and that's ad-supported, which is, in my opinion, great for the world.”

He acknowledged that the business is likely to change.

“I expect business models to evolve over time,” he said. “And it will probably still be about advertising because advertising can work better, and AI can design it better.”

Brin is confident about Google's position.

“I personally feel that as long as there is significant value being generated, we will figure out business models,” he said.

Beyond artificial intelligence, Brin was asked about the difficulties Google faces in the hardware space given recent developments in virtual reality. Google was very early in the augmented reality market with its now-defunct Google Glass.

“I feel like I made some bad decisions,” he said, referring to Google Glass. If he had done it differently, Brin said, he would have treated Google Glass as a prototype rather than a product. “But I still like the lightweight shape,” he said.

Regarding Apple Vision Pro and dead “The Quest headphones, they're very impressive,” Brin said.

When asked how Gemini might impact spatial computing or products like Google Maps or Street View, Brin responded with no less curiosity than anything else.

“To be honest, I hadn't thought about it, but now that I've said it, yeah there's no reason why we can't bring in more 3D data,” Brin said to laughter from the audience. “Maybe someone will do it in Gemini – I don’t know.”

Watch: Google vs. Google