Google’s Gemini AI image tool will be up in no time, but will it rewrite users’ prompts and insert alterations?

Google aims to relaunch Gemini AI image tool in a few weeks, a Reuters report reads.

The latest image-generating tool in the AI craze was paused some days ago, after the inaccuracy backlash over the way Gemini’s image tool generated people.

Google’s AI image generator was designed to turbo-perform on topics of diversity and inclusivity, to the point when it refused to generate white people and in historical context. The reports of the tool generating non-white Vikings went under the radar, but once Gemini AI started generating black German soldiers from the 1940s, things got really controversial.

The information for Gemini’s AI tool revival comes from Google DeepMind CEO Demis Hassabis on MWC – Google DeepMind is the unified Google AI lab DeepMind and the Google Brain (its AI research division). Last summer, it was Hassabis who announced that Gemini is coming.

Since the launch of OpenAI's ChatGPT in November 2022, Google has been racing to produce AI software to rival that of the Microsoft-backed company.

In 2023, when Bard was released (later rebranded to Gemini), it had shared inaccurate information about pictures of a planet outside the Earth's solar system in a promotional video, causing its shares to drop by as much as 9%.

“We have taken the feature offline while we fix that. We are hoping to have that back online very shortly in the next couple of weeks, few weeks”, Hassabis said in Barcelona, adding that the tool was not “working the way we intended”.

Meanwhile, people claim to know exactly what’s been going on with Google’s Gemini AI image tool: according to an X/Twitter post, Google's Gemini system seems to add phrases to a user's image-generation prompt (the instruction, such as "make a painting of the founding fathers") and inserting terms for racial and gender diversity, such as "South Asian" or "non-binary" into the prompt before it is sent to the image-generator model.

If this is correct, it means that the Gemini AI image tool is working perfectly well on the picture generating front – it’s just that it’s not your exact words it’s been fed. Does this happen with Google’s search engine as well?