Google CEO Admits AI Model ‘Biased’

Google Chief Executive Officer Sundar Pichai acknowledged that the company’s artificial intelligence (AI) model, Gemini, generated biased content, including images depicting historical figures from diverse ethnicities and genders.

This comes as he addressed the issues in an internal memo, stating that such outputs are “unacceptable” and that the company is actively working on improvements.

“I know that some of its responses have offended our users and shown bias – to be clear, that’s completely unacceptable and we got it wrong,” Pichai wrote in the memo, according to news site Semafor (Semafor)

This latest development in Google’s foray into AI model territory comes as users have shared Gemini’s image generator having historical inaccuracies, forcing the tech giant to suspend the model’s ability to generate images of people, saying that the image generator will be relaunched in the coming weeks.

AI systems are susceptible to replicating biases present in their training data.

This has been an ongoing challenge, with examples like gender bias in Google Translate and racial bias in OpenAI’s Dall-E image generation model. The recent issues with Gemini highlight the continuing complexities of mitigating bias in the development of AI. (GFB)