The Pixel 9 sequence launched final August got here with spectacular Gemini AI options, together with picture technology skills that had been too good, opening the door to abuse. It was troubling to see Google want about two months to begin fixing the Pixel Studio and Reimagine points.
These fixes, labeling Gemini-made photographs as photographs created with AI, ought to have been in place from the beginning, however they’re nonetheless not sufficient to forestall abuse.
Quick-forward to mid-March and Google has given Gemini an enormous improve, making it extra private and enhancing Deep Analysis. Gemini 2.0 Flash Considering Experimental was the mannequin Google highlighted in its announcement, the reasoning AI that makes these analysis reviews higher and helps with Gemini personalization.
It seems that Gemini 2.0 Flash can also be superb at creating and enhancing photographs. Too good. Once I say that, I imply it may be abused with ease.
Particularly, Gemini 2.0 Flash customers discovered they might take away watermarks from photographs on the internet with unimaginable ease. This basically permits any consumer to steal copyright-protected imagery from the net. It’s a lawsuit ready to occur.
You may need seen Gemini 2.0 Flash photographs on the internet through social media posts meant to spotlight the spectacular skills of Google’s image-generation instruments. You’ll discover the Gemini 2.0 Flash picture technology software in Google’s AI Studio, which is free to make use of.
Some Gemini 2.0 Flash customers discovered which you can inform the AI to take away the watermarks in pictures, and the AI will comply. The outcomes are spectacular, even with pictures that includes loads of watermarks.
For those who’ve ever purchased copyright photographs on the internet, they’re normally lined in loads of watermarks to forestall customers from cropping components of them. Gemini 2.0 Flash removes these watermarks in a, nicely, flash, as you may see within the photographs above.
The outcomes aren’t good, however they’re ok. Google most likely by no means supposed for customers to abuse Gemini 2.0 Flash this manner and take away watermarks. However Gemini does its job remarkably nicely.
Sarcastically, the AI marks the newly generated consequence with a Gemini watermark in a nook to tell individuals the picture was created with AI. That’s security measure, although one might simply crop that nook watermark out.
I’ll level out what others mentioned on Reddit: You may take away watermarks from pictures with different software program. It’s doubtless that different AI packages exist to just do that.
Nonetheless, the Gemini 2.0 Flash’s potential to take away watermarks makes the method trivial. You don’t have to show your self new expertise to take away watermarks from a picture protected by copyright. You solely have to offer Gemini a written command.
Whereas I haven’t replicated the Gemini 2.0 Flash watermark elimination function myself, others have been capable of use it. I believe it’s undoubtedly a bug that Google will repair within the not-too-distant future. I wouldn’t be stunned if future variations of AI Studio won’t take away watermarks.
I’m stunned Google didn’t take into account this downside earlier than it launched the Gemini 2.0 Flash upgrades, particularly after all of the backlash concerning the Pixel 9’s image-generation skills that lacked safeguards. Then once more, giving the AI this kind of freedom will definitely generate free publicity that is likely to be warranted within the incoming lawsuits.
