Google CEO Sundar Pichai slams ‘completely unacceptable’ errors made by Gemini AI
Sundar Pichai
Google CEO Sundar Pichai has strongly criticised the significant errors made by the Gemini AI, deeming them "completely unacceptable." These errors, which included generating historically inaccurate images such as ethnically diverse World War II Nazi troops, led to the suspension of users' ability to create people's pictures. The controversy arose shortly after Google rebranded its ChatGPT-style AI chatbot Bard as "Gemini," aiming to enhance its prominence in its products and compete with industry rivals like OpenAI and Microsoft.
In a letter to staff, Pichai acknowledged the offensive and biased responses generated by the Gemini app, saying that such shortcomings were unacceptable. He reassured that Google teams are diligently working to rectify these issues, although no specific timeline for restoring the image-generating feature was provided.
"I want to address the recent issues with problematic text and image responses in theGeminiapp," Pichai wrote in a letter to staff, which was published by the news website Semafor.
"I know that some of its responses have offended our users and shown bias -- to be clear, that'scompletelyunacceptableand we got it wrong," he added.
Challenges in AI development
Pichai acknowledged the inherent imperfections in AI technologies, particularly at this nascent stage of the industry's evolution. Despite the challenges, he affirmed Google's commitment to continuously improve its AI models, recognising the high expectations set by users.
The incident with Gemini reflects broader concerns within the tech industry regarding biases perpetuated by AI models. Companies like Google have been criticised for overlooking racial and gender biases in their AI results. Efforts to mitigate such biases have been ongoing, but challenges persist in achieving balanced and accurate outcomes.
According to Google's Prabhakar Raghavan, as cited by AFP, the issues with Gemini stemmed from attempts to address biases within the model. However, the calibration to show diverse representations sometimes led to overcompensation or excessive caution, resulting in inaccurate and embarrassing images.
Also watch |Teachers struggle to adapt amid AI revolution in education
Beyond this specific incident, the rapid advancement of AI technology raises broader concerns. Experts and governments have warned about potential economic disruptions and the spread of disinformation through AI-driven platforms, highlighting the need for responsible development and deployment of AI systems.
(With inputs from agencies)