Produce Controversial AI Images, Google Gemini Error Analysis

JAKARTA Image-making tools on Gemini chatbots have received criticism because of inaccurate results and history. As a result, this feature must be disabled.

Based on the findings of several users on platform X, the feature produces diverse images for the US Founding Fathers search and includes people with colored skin in the search for Nazi-era German soldiers.

Senior Vice President of Prabhakar Raghavan admitted that such an image result was wrong. Representing Google, Raghavan apologized and said image-producing tools at Gemini were being repaired for better results.

"It's clear that this feature missed the target. Some of the resulting images are inaccurate or even offensive. We are grateful for the user's input and apologize because this feature is not working properly," Raghavan said via Google's blog.

In fact, Google is trying to include elements of diversity to target users from around the world. However, this element of diversity cannot be controlled by Gemini's algorithm so that the image-making tool produces images that are not right.

According to Raghavan, there are two things that need to be considered from this error. First, people who use Gemini make mistakes in inputting descriptions. Like most AI tools, Gemini needs a clear range.

Second, Google just realized that Gemini is becoming more careful in producing AI images. This tool refuses to answer certain commands and misinterprets some commands it says are sensitive.

These two things lead to models providing excessive compensation in some cases and becoming too conservative in other cases, resulting in an embarrassing and incorrect picture, Raghavan explained.

Although Gemini is being repaired and will be tested again extensively, Raghavan cannot promise more accurate image results. The reason is, AI can still have flaws, but it promises secure and responsible technology.

"I can't promise that Gemini won't occasionally give embarrassing, inaccurate, or offensive results, but I can promise that we will continue to take action whenever we identify a problem," Raghavan concluded.