Google CEO Sundar Pichai calls inaccurate Gemini responses ‘unacceptable,’ promises fix

AI SaaS

Google LLC Chief Executive Officer Sundar Pichai has told employees that the company will make “structural changes” to address the errors in Gemini’s image generation feature.

Pichai detailed the plan in an internal memo distributed late Tuesday, Semafor reported today. The executive also called the error, which caused Gemini to generate inaccurate images of racially diverse people, “completely unacceptable.” Google confirmed the memo.

Gemini is the new name of the Bard chatbot the company launched last year as a response to ChatGPT. Google changed the chatbot’s name on Feb. 8 following a series of feature updates that rolled out over the preceding days. One of those updates added an image generation tool, which is what laid the groundwork for the recent Gemini controversy.

Around a week ago, users of X began reporting that some of the images generated by the chatbot inaccurately depicted racially diverse characters in historical contexts. Google responded by temporarily disabling Gemini’s ability to generate images of people. Prabhakar Raghavan, the company’s senior vice president of knowledge and information, stated last Friday that the feature will be reactivated once the issue is fixed.

Pichai shared more details on the troubleshooting effort in the Tuesday memo to employees. “Our teams have been working around the clock to address these issues,” he wrote. “We’re already seeing a substantial improvement on a wide range of prompts.”

Citing sources familiar with the matter, Bloomberg reported today that Google only equipped Gemini with the ability to generate images of people after multiple internal reviews. Nevertheless, the search giant reportedly didn’t test “all the ways that the feature might deliver unexpected results.”

Pichai detailed that Google’s effort to address the issue won’t be limited to technical troubleshooting alone. He wrote in this week’s memo that the company intends to make “structural changes” as well as roll out “updated product guidelines, improved launch processes, robust evals” and other enhancements.

The effort is also set to place an emphasis on red teaming. That’s a practice whereby a company’s engineers attempt to find user output, such as chatbot prompts, with the potential to cause application malfunctions. Last July, Google disclosed that it has formed a team focused specifically on finding issues in its AI products.

Gemini is based on a large language model family of the same name that was developed by the search giant’s Google DeepMind unit. The unit also created the Imagen 2 model that powers the chatbot’s image generation features. Demis Hassabis, DeepMind’s CEO, disclosed on Monday Gemini’s ability to generate images of people could be reactivated within a few weeks.

Image: Google

Your vote of support is important to us and it helps us keep the content FREE.

One click below supports our mission to provide free, deep, and relevant content.  

Join our community on YouTube

Join the community that includes more than 15,000 #CubeAlumni experts, including Amazon.com CEO Andy Jassy, Dell Technologies founder and CEO Michael Dell, Intel CEO Pat Gelsinger, and many more luminaries and experts.

“TheCUBE is an important partner to the industry. You guys really are a part of our events and we really appreciate you coming and I know people appreciate the content you create as well” – Andy Jassy

THANK YOU

AI SaaS

Leave a Reply

Your email address will not be published. Required fields are marked *