Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

MacRumors

macrumors bot
Original poster
Apr 12, 2001
67,643
38,064


Google last week introduced the Pixel 9 smartphone lineup, with a selection of new devices that are focused on AI. All of the Pixel 9 phones support Gemini AI, and Google has also added AI-based image generation and editing tools. Reviewers have been testing the new smartphone features, and the AI image generation is probably going to turn into a PR nightmare for Google based on what's come out so far.

google-pixel-9.jpg

Google introduced an AI image generation app for the Pixel phones, called Pixel Studio. It's designed to create stickers and images using text-based prompts, and it looks and sounds a lot like Apple's planned Image Playground. So far, reviewers have been able to create all manner of questionable images with Pixel Studio, ranging from a Nazi version of Spongebob to Elmo with an AK47.

pixel-studio-digital-trends.jpg
Image via Digital Trends​

Digital Trends had no trouble getting Pixel Studio to make images of beloved cartoon characters doing questionable activities. Guns, drugs, and alcohol were not off-limits, nor were offensive situations like school shootings. Pixel Studio did not generate these kinds of images unless asked, but that's most likely what people will immediately do with the feature rather than generating images of cute cats and bunnies.

Google said that there were "safety checks" in place to keep Pixel Studio from being "used nefariously," and there are some. Pixel Studio will not create images of humans, and Google is tweaking the AI in response to reviews. After speaking with Google, Digital Trends was no longer able to get Pixel Studio to create cartoon characters using cocaine or dressed as German soldiers.

Even more concerning than Pixel Studio is Google's "Reimagine" tool, which can add objects to photos that you've already taken. The Verge used it to add corpses, bombs, drugs, and disasters to images, and objects included in photos look so realistic that it's difficult to tell when an image has been edited. Google is able to do a pretty good job matching lighting and perspective to the original photo, and there are no watermarks or flags on social media. Google does add a metadata tag, but that's easy enough to eliminate with a screenshot.

You can, of course, use Reimagine for adding sunsets and rainbows to your image, just like you can use Pixel Studio for creating fun pictures, and both AI features work great, so these tools aren't all bad. Android Authority shared a post with Oreo pizza, a giraffe surfing, and kittens playing basketball, for example. As The Verge points out, you could always add a body to an image or create a picture of Elmo with a gun, but it would take some Photoshop skill and time to do so. With the Pixel phones, it takes just a thought and a few seconds to create that kind of image, and it's right there on the smartphone for anyone to use.

In statements to The Verge and Digital Trends, Google said that it designed its generative AI tools to "respect the intent of user prompts" which can lead to the creation of content "that may offend" when the user asks for it. But Google claims that there are "Terms of Service" on what content is not allowed, and that safeguards will be continually refined.

Apple hasn't yet introduced its image generation tools, Image Playground or Genmoji, and the media surrounding the Pixel 9 launch gives us some insight into what the response might be like if Apple doesn't get it just right. With AI, it's likely that people are going to be able to find workarounds even if Apple puts numerous safeguards in place. Apple so far has shied away from photorealistic AI images, but Google is going all in, and it's going to be much harder to trust photos on social media and the internet going forward.

Article Link: Google's Pixel AI Image Creation and Editing Tools Are Kind of Terrifying
 
How many times will it take before Google and other tech giants learn that A) we don’t want this and B) no matter how much you try, people WILL find a way to use your products for nefarious purposes. They’re going to be playing whack-a-mole with their algorithm until it’s nerfed to the point it can only generate images of flowers, baby animals, and the latest product from their respective hardware branch.


IMG_0136.jpeg
How many times do we have to teach you this lesson, old man?
 
Guns, drugs, and alcohol were not off-limits, nor were offensive situations like school shootings. Pixel Studio did not generate these kinds of images unless asked, but that's most likely what people will immediately do with the feature rather than generating images of cute cats and bunnies.

Really? Might be normal behavior in one certain country, but definitely not worldwide.
 
A new era of freedom of speech (protected speech) has been ushered in by the advent of easily accessible AI. What is “nefarious”? Should megacorps who operate above the law be the sole judge of what is acceptable? Will the powers that be have unrestricted access to generating any image that their hearts desire? Unfortunately, 90% of the population could not care less about these questions.

Training your children to be critical thinkers will be paramount to preserving the truth.

The war of information rages on…
 
“Oh no! The computer’s doing what I asked it to!” is something I’ll never get. As long as they don’t allow it to generate realistic images featuring real people it’s whatever because nothing’s stopping anybody from picking up Procreate and drawing these anyway.
 
“Oh no! The computer’s doing what I asked it to!” is something I’ll never get. As long as they don’t allow it to generate realistic images featuring real people it’s whatever because nothing’s stopping anybody from picking up Procreate and drawing these anyway.
"Even more concerning than Pixel Studio is Google's "Reimagine" tool, which can add objects to photos that you've already taken. The Verge used it to add corpses, bombs, drugs, and disasters to images, and objects included in photos look so realistic that it's difficult to tell when an image has been edited. Google is able to do a pretty good job matching lighting and perspective to the original photo, and there are no watermarks or flags on social media. Google does add a metadata tag, but that's easy enough to eliminate with a screenshot."
 
How many times will it take before Google and other tech giants learn that A) we don’t want this and B) no matter how much you try, people WILL find a way to use your products for nefarious purposes. They’re going to be playing whack-a-mole with their algorithm until it’s nerfed to the point it can only generate images of flowers, baby animals, and the latest product from their respective hardware branch.


View attachment 2408149
How many times do we have to teach you this lesson, old man?
While I don’t want the generated images in the article, I want AI image generators. I use them for some of my university course presentations when I could use a particular image that’s not readily available anywhere. I don't have available funds to pay for a stock image or pay to have someone create me an image. I also use them to quickly mock up ideas for things.

Edit: I think it's funny that people disagree with me saying I find AI image generators useful. Your disagreement (or agreement for that matter) doesn't change the fact that I find them useful.
 
Last edited:
I can see the potential for copyright complaints, but other than that we seem to melt down over the silliest things these days...

It's dumb, frivolous, unproductive and lazy-- but terrifying?

I think the realistic ones where The Verge was able to add corpses, drugs, etc. are pretty concerning. Paddington Bear on the cross less so, but still not ideal.
 
How many times will it take before Google and other tech giants learn that A) we don’t want this and B) no matter how much you try, people WILL find a way to use your products for nefarious purposes. They’re going to be playing whack-a-mole with their algorithm until it’s nerfed to the point it can only generate images of flowers, baby animals, and the latest product from their respective hardware branch.

A and B don't go together. There are lots of people who "want this" and, in my opinion, there are valid use cases too. Whether those use cases should be served by a tool that's unrestricted and available to the masses is a separate question entirely.
 
Terrifying, how?

And stores sell these things called knives which are very dangerous. And cliffs exist. And boxes of pills. And the sea is too rough. And old age is lethal. Cars have crashed. People have fallen from ladders and broken legs. Soccer tackles! Fists can fly too quickly. Let’s all get back under a blanket… just not the same blanket. Call me when it’s all over and Elmo goes to AA.
 
Last edited:
well, at least no-one seems to have tried to show Sponge Bob eating rocks, and using glue on his pizza ...
oh wait, that was openai, wasn't it?
Actually, those were Google AI as well 🤦‍♂️


Google's new artificial intelligence (AI) search feature is facing criticism for providing erratic, inaccurate answers.

Its experimental "AI Overviews" tool has told some users searching for how to make cheese stick to pizza better that they could use "non-toxic glue".

The search engine's AI-generated responses have also said geologists recommend humans eat one rock per day.
 
  • Like
Reactions: KeithBN
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.