Hi, this is Wayne again with a topic “THIS is Google’s SOLUTION for detecting AI fakes!”.
I’M going to show you a photo. I took using my pixel 7 Pro. I shot this at Devil’s Hopyard State Park, which is a popular place for photographers to go in Connecticut. It’S a great photo right. Well, it would be if that was the actual photo that I shot that day. What am I talking about? Let’S dive in all right, so you saw the photo I captured at Devil’s Hopyard State Park. However, that isn’t the real photo, I captured technically it’s the real photo, but it’s the photo that came out after I fed it to Magic editor, which is an AI based tool that Google introduced this year.
This is the actual photo. I shot that day. Take a look at them side by side, and you can see that the magic editor version looks drastically different from the original version. Now it might not seem like it, but this is a big problem today.
I want to go over why this is a problem. What Google is and isn’t doing about the problem and things that we think might be able to solve this problem? So so, first, let’s talk about how this photo was made. First, I took the photo using my pixel 7 Pro. Then I fed it to Magic editor.
Magic editor was launched earlier this year by Google at Google iio. Essentially magic editor is like having Photoshop in your pocket, except that you don’t have direct control over the edits that it’s making. Instead, you feed it the image it interprets that image and then tries to make that image better using the power of generative AI. The problem, though, is that, when you rely on AI to edit your images for you you’re, eventually going to end up with some sort of distortion of reality.
In the case of this Devil’s Hopyard photo. What came out doesn’t look like the actual environment that I was in when I shot it. In fact, I would say that this photo is a fake photo and I’m the one who took it when Google announced magic editor. It admitted this kind of thing could be a problem. The solution that it came up with was a watermarking system. Basically, when you take a photo that you’ve captured and put it through magic editor, it has a watermark attached to it that lets you know.
Magic editor has manipulated this photo. Google said that this system would make it easy for anyone to find out whether or not a photo was actually a real photo or if it had been manipulated with AI. Naturally, the first thing we did was wonder how to see this Watermark.
We tried all the obvious ways, but we couldn’t figure it out. So we reached out to Google to find out what we need to do. Google got right back to us and told us that, in order to see The Watermark, we need to upload the photo to the international press.
Telecommunications Council website or IPC IPC has a website where you can upload an image or a URL, and it will process that image and tell you the metadata inside it sure enough. When we took our magic Editor Photo. We put it through the IPC tool and it came out with a tag that said AI generated with Google.
We took the same photo that I did not put through magic, editor put that through the IPC tool and sure enough. Their tag was not there. Therefore, this tool can easily distinguish between a photo, that’s been manipulated with magic, editor and the same photo not manipulated through magic.
Editor problem solved right. Well, not exactly. We see a number of issues with this system. The first is very obvious, which is no one is going to visit this website.
I didn’t even know this website existed until we asked Google about it. So surely everyone else in the world who just is a normal, smartphone user, isn’t going to know it’s there either. Why does this have to be so cumbersome? Why do they have to visit a third party website in the first place shouldn’t this just be baked into Google photos, or maybe even Google, Chrome or Gmail or any other of the Google apps out there? Another issue we see is the security around the watermark itself. What’S to stop someone from downloading the magic Editor Photo, removing the metadata and then re-uploading it to the internet.
Then, when someone uses this IPC tool on that magic Editor Photo, the tool will tell them that it has not been manipulated by Magic editor. This, of course, creates a whole host of issues any other AI based system you use to edit. Your photos isn’t going to be picked. Pied up by the IPC tool like a magic Editor Photo will be so this creates a whole swarm of problems, because now you have an AI generated photo that you upload to the IPC tool and it spits it back out and says there is no watermark on This so it’s totally fine.
These are a lot of problems and we need to know what Google’s going to do about them. Now, if you’ll remember, Google was the one who told us about the IPC tool in the first place and they told us about it very quickly after we asked. So we responded to that with a whole host of other questions, asking for more clarity on. What’S exactly going on here, unfortunately, Google didn’t get back to us.
We sent Google a reminder, email and it still didn’t get back to us. So unfortunately, we don’t know any of the answers to these questions that we’re presenting all we know is that there’s an IPC tool that can check magic, editor photos, that’s it now. I am not an expert about AI generative imagery, but even I can tell that there are a whole swarm of issues related to this policy that Google has right now. Magic editor as it stands, isn’t entirely safe to use.
Yes, you can use it and it can be fun to change your photos and make them look all different and weird, but you’re going to have problems where you’re distorting the reality of what’s going on. And if you upload that image to the internet, it could spread. Misinformation about what that photo is captured to counteract this. There shouldn’t be a website that you have to visit to then manually upload a photo to check whether or not it’s real.
It has to be easier than that with Google’s wide ranging ecosystem of products, including once again, Gmail, Chrome, Android, Etc. It should be a lot easier to figure out out whether or not a photo has been AI generated or Not. Another concerning thing about this whole system is that there doesn’t appear to be any limitations set on any other companies that are working with Google that are also doing AI generated imagery. For example, we know that Samsung is going all in on AI, for the Galaxy s24 series which comes out early next year does Samsung have to follow the same watermarking rule. In other words, if I create an image using Samsung’s, AI generative system and then upload that to the IPC tool is, is that going to also come back with a similar Watermark? If it does that’s great, but if it doesn’t, what does that mean? Is there going to be a separate tool? This is going to be way too confusing, and this needs to be streamlined and Google is in the most powerful position to do that. The way we see it, Google is one of the richest most powerful companies in the tech sector, and it is uniquely positioned to be able to Institute rules around AI generative imagery and enforce them, with over 3 billion Android devices out there, and over 3 billion Chrome Users, no other company, is poised in the proper place to make sure that AI generative imagery stays ethical, look I’ll, admit. Magic. Editor is a lot of fun. It’S cool to take a photo upload it to Magic, editor and see the wacky things that come out.
I really enjoyed testing it out. However, I do see the ethical problems here. Ai is changing our world at such a rapid pace and if we don’t keep up we’re going to fall behind, Google had no issue rolling out magic, editor very quickly, but seems to be dragging its feet when it comes to making sure that there are safeguards surrounding Magic Editor to prevent it from being used for misinformation purposes.
The company also doesn’t seem to have that much to say about how it’s going to enforce ethical standards on other companies that it works with. When it comes to these kinds of systems, our world is already filled with so much misinformation and tools like magic editor have the potential to make that problem even worse, and the only way we’re going to stop. That is, if company like Google step in and prevent it from happening, so Google, if you’re watching this, please make sure that you’re, not just spending all of your time. Making these cool new tools make sure you’re spending just as much time making sure that these cool new tools are safe for more information on this topic be sure to check the link in the description.
But before you go make sure you go to the comments and let us know what you think about magic editor and AI generative imagery in general. I’Ll see you in the next article .