Google Explains Why Gemini Creates Nonsensical Images

In a news we shared with you recently, we mentioned that Google Gemini’s visual creation tool started to give extremely politically correct results. Immediately after Googlemade a statement on the issue and first apologized for these erroneous visual results and then announced that it had disabled the visual creation tool.

Sharing a new blog post, Google explained why its image creation tool creates extremely politically correct images. According to Google’s statement Gemini, first misunderstood Google’s ‘intent’ and then rejected the corrections it had made to address political correctness. As a result, the image generator had to be disabled until it was fixed.

Gemini chose to draw by correcting users’ racist and sexist(!) demands

Gemini’s result when asked to draw “A German woman”.

According to Google’s statement, Gemini is a special project independent of the search engine and other artificial intelligence products, and its image creation feature is built on an artificial intelligence model called Imagen 2. Google trained Gemini to give the most realistic results when commands like “football player” or “someone walking their dog” were entered. educates to include different genders and races states.

The problem is that Gemini also works in commands that have historical or racial contexts. to anti-discrimination continued. The worst part is that Gemini became more “cautious” with use and rejected Google’s efforts to fix it. So when you asked him to draw a “medieval English king” he drew a black female king. Because, according to Gemini, expecting it to draw a white man when you gave this command was a racist and sexist expectation.

What will happen next?

Gemini’s conclusion when asked to draw “an English king in the Middle Ages”.

Google underlines that it does not name Gemini’s refusal to draw certain groups, races, genders. He states that they will make serious corrections and comprehensive tests before launching the visual creation tool. However, Google said that even after corrections, artificial intelligence still has problems during image creation. It’s normal to make mistakesHe stated that the development process will continue.

Google senior vice president Prabhakar Raghavan clearly states,Gemini’s occasional I can’t promise it won’t produce embarrassing, inaccurate or disturbing resultsWhen using the expressions “But I can promise that we will continue to take action when we detect a problem.” says.