Google CEO Sundar Pichai comments on multimodal AI Gemini's excessive diversity problem as ``totally unacceptable''

by World Economic Forum

Google's multimodal AI 'Gemini' outputs completely incorrect images due to too much consideration for racial diversity, CEO Sundar Pichai said, 'It is completely unacceptable' and addressed the issue. He said he would work around the clock to fix it.

Google CEO Sundar Pichai calls AI tool's responses 'completely unacceptable' | Semafor
https://www.semafor.com/article/02/27/2024/google-ceo-sundar-pichai-calls-ai-tools-responses-completely-unacceptable

Google CEO: Gemini AI photo diversity scandal 'offended our users' - The Verge
https://www.theverge.com/2024/2/28/24085445/google-ceo-gemini-ai-diversity-scandal-employee-memo

This problem is caused by the fact that when Gemini outputs images of Nazi German soldiers, it generates images of blacks and Asians, or rejects depictions of white people, resulting in 'historical It was criticized as saying, ``It may not be possible to obtain accurate results.''

Google apologizes for Gemini generating 'racially diverse Nazis' - GIGAZINE

Google explained the cause by saying, ``As a result of emphasizing diversity, we were unable to respond to cases where we should not do so'' and ``Gemini began rejecting certain prompts,'' and turned off the image generation function for people. announced that it would make major improvements.

Google explains why a mistake occurred in image generation with multimodal AI 'Gemini' - GIGAZINE

'We would like to address issues with Gemini's text and image responses,' Pichai said in a memo obtained by online news outlet Samafor. 'I understand. Let me be clear: this is completely unacceptable and we made the wrong decision.'

CEO Pichai added, 'Fixes to Gemini are already underway and we are working around the clock. We are already seeing significant improvements across a variety of prompts. Especially at this new stage in the industry's development, There is no such thing as a perfect AI, and we know that the hurdles are high for us, but we will keep trying no matter how long it takes.We will also review what happened and fix it at scale. We will continue to do so.”

'We know what it takes to make great products that billions of people and businesses love,' Pichai said in the memo. , it gives us a foothold to ride the big wave of AI. Let's think about what we should focus on most, and that is to create useful products that users can trust.'