Pike survived a lethal injection execution on Wednesday. She was sentenced to death for the 1995 killing of her classmate, Colleen Slemmer.
Christa Pike, the woman who survived an execution attempt in Tennessee earlier this week, is on a ventilator as she receives care in the hospital, her lawyers said in a new statement Friday.
“Christa remains critically ill and is receiving medical care for the devastating effects of Wednesday night’s events,” the lawyers’ statement said. “She is critically ill, unconscious, intubated and receiving breathing assistance through a ventilator.”
The statement said she is being treated for “the serious effects of the pentobarbital administered during the execution attempt, including significant injuries to both arms.”
Forensic analysis can already do that. Yes, even AI tampering can be proven.
Being able to prove a positive doesn’t imply being able to prove a negative.
Except you wouldn’t need to prove a negative. If someone claims the evidence is AI generated, they have to prove that, not you that it isn’t AI generated.
And that’s what forensic analysis does - it looks at the video (incl metadata, pixel level analysis, and so on), compares other, known good evidence (say, the attestation of the sensor, or, as I mentioned, the sensor-unique pixel patterning), and confirm the authenticity by that.
Just because you’re unable to comprehend how thorough such analysis can be, and how well AI artifacts can be detectef if you really look at it, doesn’t mean that scientists whose day job is literally doing that, don’t have the methods.
And then we haven’t even talked about e.g. SynthID, something that I suspect will soon be regulation to be embedded into any visual genAI model.
I feels like you’re missing the point. There’s a tons of cases where innocent people get prison time despite no hard proof on them(via coercion or the lack of alibi), then there’s guilty people who hired a gang of lawyers to throw out hard evidence one by one(by finding every minor mistake by the prosecutor). It’s incredibly hard to proof people are beyond reasonable doubt that they should be executed, especially when the guilty can just play innocent.
How?
“trust me bro”
Ask a forensic analyst? I am not one so my knowledge of it is cursory at best.
However recent hardware now includes hardware level attestation in the image files that cannot be reproduced, so validity of video and photo evidence can NOW be established easily. If the signature is there, easy, the video wasn’t manipulated. If it is missing, the evidence is no longer attested, therefore questionable.
How do you know the attestation key hasn’t been compromised, allowing me to sign any video I want?
Because the attestation isn’t single step, isn’t a singular certification/signature.
Go read up on how Google and Apple are handling it.
Can you mathematically prove it’s impossible?
Can you actually take the effort to understand how attestation works, so you can realise just how dumb your question is?
For now. At this point, it doesn’t matter much if the AI economic bubble collapses, research will continue, potentially at a lesser pace, and could still outpace, even if temporarily, detection methods.
Yeah, uh, no.
It might avoid free eye detection (still doesn’t, AI content is still incredibly easy to spot), but FORENSIC analysis is a bit more thorough than that.
Though I really love these kinds of comments. I at least used to work in the field some 15 years ago, so I have some awareness of what’s going on. Meanwhile the group of people here who saw flashy AI demos and two-three episodes of CSI think they know best.
For example, here’s a fun fact: camera sensor pixels aren’t equal. Even on the same sensor, there’s a subtle level of difference in sensitivity between pixels, pixel blocks, and so on. The built in ISP also has its own unique signature. Hell, there’s even a unique signature in the variations of the variance between key frames! All of which can be extracted - much like a hash - but would take immense computational power (we’re talking quantum computing levels, since it IS essentially reversing a hash or cryptographic key) to properly break.
This kind of fingerprinting alone could verify if a video was tampered with nor not. There’s been whitepapers about this kind of verification since at least 2012. Yet I bet none of you who oh so confidently downvote, knew about it.
Humans will still be behind the “forensic science”, and humans are fallible. There was just a DNA forensic analyst who was found guilty of evidence tampering. She was overworked so just… Made stuff up.
https://www.nbcnews.com/news/us-news/colorado-crime-analyst-manipulated-dna-data-sentenced-rcna596722
Also as someone who currently works in the legal field, expert witness testimony is not the slam dunk you assume it is.
White papers aside has it been done in court yet?