Google's AI Detection Tool Can't Decide If Its Own AI Made Doctored Photo of Crying Activist.
When the White House X account posted an image depicting activist Nekima Levy Armstrong in tears during her arrest, there were telltale signs that the image had been altered. Less than an hour before, Homeland Security Secretary Kristi Noem had posted a photo of the exact same scene, but in Noem’s version Levy Armstrong appeared composed, not crying in the least.
To determine if the White House version of the photo had been altered using artificial intelligence tools, researchers turned to Google's SynthID - a detection mechanism that Google claims is able to discern whether an image or video was generated using Google's own AI. They followed Google's instructions and used its AI chatbot, Gemini, to see if the image contained SynthID forensic markers.
The results were clear: The White House image had been manipulated with Google’s AI. Researchers published a story about it.
However, after posting the article, subsequent attempts to use Gemini to authenticate the image with SynthID produced different outcomes. In their second test, Gemini concluded that the image of Levy Armstrong crying was actually authentic.
But in their third test, SynthID determined that the image was not made with Google's AI, directly contradicting its first response. This inconsistency raises serious questions about SynthID’s reliability to tell fact from fiction at a time when AI-manipulated photos and videos are growing increasingly prevalent.
Aside from Google's proprietary tool, there is no easy way for users to test whether an image contains a SynthID watermark. That makes it difficult in this case to determine whether Google's system initially detected the presence of a SynthID watermark in an image without one or if subsequent tests missed a SynthID watermark in an image that actually contains one.
If AI-detection technology fails to produce consistent responses, though, there’s reason to wonder who will call bullshit on the bullshit detector. The incident highlights concerns about the reliability of AI-generated tools and their potential misuse in spreading misinformation.
Supporters of the technology argue tools that can detect if something is AI will play a critical role establishing the common truth amid the pending flood of media generated or manipulated by AI. However, if AI-detection technology fails to deliver accurate results, the very notion of fact-checking becomes meaningless.
When the White House X account posted an image depicting activist Nekima Levy Armstrong in tears during her arrest, there were telltale signs that the image had been altered. Less than an hour before, Homeland Security Secretary Kristi Noem had posted a photo of the exact same scene, but in Noem’s version Levy Armstrong appeared composed, not crying in the least.
To determine if the White House version of the photo had been altered using artificial intelligence tools, researchers turned to Google's SynthID - a detection mechanism that Google claims is able to discern whether an image or video was generated using Google's own AI. They followed Google's instructions and used its AI chatbot, Gemini, to see if the image contained SynthID forensic markers.
The results were clear: The White House image had been manipulated with Google’s AI. Researchers published a story about it.
However, after posting the article, subsequent attempts to use Gemini to authenticate the image with SynthID produced different outcomes. In their second test, Gemini concluded that the image of Levy Armstrong crying was actually authentic.
But in their third test, SynthID determined that the image was not made with Google's AI, directly contradicting its first response. This inconsistency raises serious questions about SynthID’s reliability to tell fact from fiction at a time when AI-manipulated photos and videos are growing increasingly prevalent.
Aside from Google's proprietary tool, there is no easy way for users to test whether an image contains a SynthID watermark. That makes it difficult in this case to determine whether Google's system initially detected the presence of a SynthID watermark in an image without one or if subsequent tests missed a SynthID watermark in an image that actually contains one.
If AI-detection technology fails to produce consistent responses, though, there’s reason to wonder who will call bullshit on the bullshit detector. The incident highlights concerns about the reliability of AI-generated tools and their potential misuse in spreading misinformation.
Supporters of the technology argue tools that can detect if something is AI will play a critical role establishing the common truth amid the pending flood of media generated or manipulated by AI. However, if AI-detection technology fails to deliver accurate results, the very notion of fact-checking becomes meaningless.
This whole thing is wild 
. I mean, who creates a tool that can't even tell if it created the image itself? 
It's like me saying my clothes are clean when they're actually stained with last night's pizza sauce... not gonna fly 
. It just goes to show how flawed AI systems can be and how hard it is to trust them with important tasks like fact-checking. What's next, a lie detector that keeps telling you the truth? 
- like SynthID from Google. If it can't even decide whether its own AI made a doctored photo, how can we trust it in real life? 
I mean, if Gemini said one thing and SynthID said another, that's just too much to handle. What's next? Using AI to fact-check the AI detectors? 
That's like creating a never-ending loop of BS 
It's time for Google (and us) to get this under control! 
if we can't trust our own AI-detection tools, how can we trust anything else? this is a serious problem that needs to be addressed ASAP 
it's like the old saying "if you can't trust the source, don't trust the info" 
anyway, this is a major concern for me rn. who's gonna fact-check if AI-detection tools are failing us? 
It's not like we should be worried about misinformation spreading or anything... I guess it's good that Google is trying to develop this tech, but wouldn't it be more effective if they just came clean and said "hey, our tool can get it wrong sometimes"? 
And we need to be cautious not to get caught in the echo chamber of our own biases... 
I mean, how can we expect to trust AI-generated "fact-checking" if we're just going to rely on its own flawed logic? 
but what if we can't trust the tool that's supposed to detect them? 
. If AI-detection technology can't get its act together, what's the point of even having fact-checking? 
. The whole thing with the White House image of Nekima Levy Armstrong crying is just wild - one test says she's faking it, and the next one says nope, she's actually got emotional trauma 
. It's hard to know what's real when these tools can't even decide for themselves! And if they're not accurate, who's going to fact-check for us? 
just because we have fancy tech doesn't mean our tools are infallible. AI-detection is like the ultimate test of human skepticism - if we can't trust these tools, how can we trust ourselves to discern truth from lies? it's all about perspective and accountability 
.
This whole thing is like, totally flawed, you feel me? 
. But here's the thing, I started thinking about how those desserts were probably made with AI-generated designs or 3D printing technology... it got me wondering, are we already living in a world where AI is being used to create everything from art to food? It just blows my mind!
first off, Google's AI detection tool can't even trust its own tech, that's just crazy talk 
idk but I'm not buying it just yet 
.
It's like, we're already living in a sci-fi movie or something. But seriously, someone needs to figure out how to make these tools reliable ASAP.
. Otherwise, we're just perpetuating the problem and spreading misinformation left and right