News Technology Here's what Sundar Pichai said on Gemini AI inaccurate image generation

Here's what Sundar Pichai said on Gemini AI inaccurate image generation

Google CEO Sundar Pichai in an interview acknowledged that the company made a mistake when it generated racially biassed images depicting Nazis and the US Founding Fathers.

Sundar Pichai Image Source : PTISundar Pichai

Google recently introduced a new image generation feature on its chatbot, Gemini (formerly known as Bard), which could create images of people. However, the feature failed to meet the company's expectations and generated some inaccurate and offensive images, leading to the company pausing the feature altogether.

During an interview with Bloomberg, Google CEO, Sundar Pichai, admitted that the company had made a mistake earlier this year when it generated racially biased results in response to user queries. Pichai acknowledged that the company's intentions were good, but the outcome was not what was expected. As a result, Google will be re-releasing the image generation capabilities in Google Gemini in a few weeks. 

The company had quickly paused Gemini's image-generating feature after social media users complained that it was creating 'wrong' images. One image that caused much uproar depicted Nazis and the US Founding Fathers, and included people of colour. Pichai termed the AI tool's controversial responses 'completely unacceptable' and ordered a complete rebuild.

According to Pichai, they are retraining the models of Gemini from scratch to improve the product. He stated that as soon as the product is ready, it will be released to the public. Previously, Pichai had said that the answers provided by Gemini are consistent with the answers that are available on Google Search. 

Pichai further added that they aim to provide trustworthy and high-quality information to users. He emphasised that while it is important to provide an answer to a question, sometimes it is necessary to provide a range of opinions from the web, which is what the search does well. 

Pichai explained that the summary provided by Gemini is not contradictory to these principles, as it still aims to provide a range of opinions. The company has acknowledged the problem and attributed it to the limitations in the training data used to develop Gemini. Even Sergey Brin, the co-founder of Google, admitted that the AI chatbot "definitely messed up" while speaking at San Francisco's AGI House.

ALSO READ: Neuralink brain chip implant completes 100 days: Here's how first participant is performing