Minichan

Topic: The dangers of AI

Anonymous A started this discussion 2 years ago #115,645

What will we do if we are unable to trust video evidence any more?

Probably the same thing people did for thousands of years before we had any way to record video or audio.

The brief period that we could've vaguely trusted video/audio evidence wasn't even as useful as it could've been because a lot of places had all-party-consent laws. The people that need to be recorded and held to reality the most, such as bosses or other authority figures, simply never consent to being recorded. Or if they do, they change their behaviour as soon as the mic is off. And then if anything ever goes to court, it's a he said vs she said type of situation, where courts will typically side with the person that already has the most authority/highest status in society. One party consent was one of the few times where this playing field was levelled.

Anyway. Doesn't matter because we were all too stupid to really make full use of it anyway while we could.

/rant

Anonymous B joined in and replied with this 2 years ago, 1 minute later[^] [v] #1,277,506

> thousands of years
> video

Anonymous A (OP) replied with this 2 years ago, 3 minutes later, 4 minutes after the original post[^] [v] #1,277,507

@previous (B)
You need to work on your reading comprehension.

Anonymous B replied with this 2 years ago, 49 seconds later, 5 minutes after the original post[^] [v] #1,277,508

@previous (A)
You need to work on your concept of human history. Thanks.

Anonymous A (OP) replied with this 2 years ago, 3 minutes later, 9 minutes after the original post[^] [v] #1,277,510

@previous (B)
Sorry, you're too dumb to discuss this subject with. But enjoy this image of a heart: ♥️ as my way of sending a digital token of something to you. Enjoy!

Anonymous B replied with this 2 years ago, 4 minutes later, 14 minutes after the original post[^] [v] #1,277,513

@previous (A)
Stop changing the subject and. learn human history. Thanks.

boof joined in and replied with this 2 years ago, 16 minutes later, 30 minutes after the original post[^] [v] #1,277,521

ok fine, trillions of years it is

Anonymous D joined in and replied with this 2 years ago, 7 hours later, 8 hours after the original post[^] [v] #1,277,554

Just ask AI to design a camera that signs videos with unfalsifiable timestamps using radio entropy or something. People have been thinking about this for a while now, and that's one of the ideas to solve the problem. It's actually been a problem for a long time, regardless of AI

(Edited 3 minutes later.)

Anonymous D double-posted this 2 years ago, 2 minutes later, 8 hours after the original post[^] [v] #1,277,555

Then you can bring a cryptography expert to court to testify that this video was signed in the camera using a key owned by Panasonic or whatever camera company and the timestamp proves it was recorded when it says it was

(Edited 1 minute later.)

Anonymous E joined in and replied with this 2 years ago, 2 minutes later, 8 hours after the original post[^] [v] #1,277,557

@1,277,506 (B)

Learn to read dummy

Anonymous F joined in and replied with this 2 years ago, 45 minutes later, 9 hours after the original post[^] [v] #1,277,566

Who's Al?

Meta !Sober//iZs joined in and replied with this 2 years ago, 43 minutes later, 10 hours after the original post[^] [v] #1,277,570

It'll just be like the thousands of years where we didn't have perfectly true recordings of things.

It won't be anything new really.

Anonymous A (OP) replied with this 2 years ago, 2 hours later, 12 hours after the original post[^] [v] #1,277,606

@previous (Meta !Sober//iZs)
That's what I was saying, yes.

And to add, the level playing field that we could've had with the benefit of one-party consent laws will be gone. But we, as a society, were too stupid to fully benefit from that anyway, so no real loss.

Anonymous A (OP) double-posted this 2 years ago, 9 minutes later, 12 hours after the original post[^] [v] #1,277,608

@1,277,554 (D)

> Just ask AI to design a camera that signs videos with unfalsifiable timestamps using radio entropy or something. People have been thinking about this for a while now, and that's one of the ideas to solve the problem. It's actually been a problem for a long time, regardless of AI
> Then you can bring a cryptography expert to court to testify that this video was signed in the camera using a key owned by Panasonic or whatever camera company and the timestamp proves it was recorded when it says it was
Not sure what radio entropy has to do with anything, but digitally signing would definitely work at least on a technical level. Seems a bit Minister of Truth tier though and doesn't really put us back to where we were for the overwhelmingly brief period of time where we had video and audio recording but no AI. Of course videos could be faked before AI, but in many situations it would've been so exceedingly expensive that any reasonable person could've concluded that, say in a court room with regular people, neither party would've had the resources to do so. This never required the trustworthiness of a 3rd party, like your suggestion, and wouldn't be enough to convince someone if Panasonic or the Minister of Digital Truth Signing itself were under question.

Ultimately it just comes down to the fact that we will end up having to take things on trust. Digitally signing stuff doesn't fundamentally change that. But as we've both been saying, it's really nothing new in total.

Anonymous A (OP) triple-posted this 2 years ago, 7 minutes later, 12 hours after the original post[^] [v] #1,277,611

@1,277,521 (boof)
There are more people on earth than atoms in the universe.

boof replied with this 2 years ago, 6 minutes later, 12 hours after the original post[^] [v] #1,277,613

@previous (A)
FACTS

Meta !Sober//iZs replied with this 2 years ago, 12 minutes later, 12 hours after the original post[^] [v] #1,277,615

@1,277,606 (A)
Personally I look forward to the day any kid can generate 4k video of Trump saying the N word or Hillary eating adrenochrome from babies (or whatever the hot button issues are 20 years from now).

Anonymous A (OP) replied with this 2 years ago, 18 minutes later, 13 hours after the original post[^] [v] #1,277,625

@previous (Meta !Sober//iZs)
It'll be possible in 10 years (probably more like 5). By 20 years it'll be so old hat that I'm not sure anyone will bother.

Anonymous D replied with this 2 years ago, 3 hours later, 16 hours after the original post[^] [v] #1,277,697

@1,277,608 (A)
Through cryptography you could be 100% certain that Panasonic claims the video came from one of the cameras they manufactured. That's not trust, that can be a verifiable fact if they create that system.

The radio entropy shit was a bit shitposty but I was thinking about the timestamping aspect. Trustless timestamps using atomic decay, but I guess they could use just a blockchain lol

(Edited 3 minutes later.)

Anonymous D double-posted this 2 years ago, 4 minutes later, 16 hours after the original post[^] [v] #1,277,698

@1,277,615 (Meta !Sober//iZs)
I'm going to wear Trump's Metaverse avatar and do a shit on that guys potatoes

Anonymous A (OP) replied with this 2 years ago, 11 minutes later, 16 hours after the original post[^] [v] #1,277,702

@1,277,697 (D)
> Through cryptography you could be 100% certain that Panasonic claims the video came from one of the cameras they manufactured. That's not trust, that can be a verifiable fact if they create that system.
Yes, it requires you to trust Panasonic is: A) keeping their private keys secure, B) not intentionally fabricating evidence.

Anonymous D replied with this 2 years ago, 3 minutes later, 17 hours after the original post[^] [v] #1,277,704

@previous (A)
You don't have to trust panasonic. You will KNOW that panasonic says the video came from one of their cameras

Anonymous A (OP) replied with this 2 years ago, 32 minutes later, 17 hours after the original post[^] [v] #1,277,709

@previous (D)
> You will KNOW that panasonic says the video came from one of their cameras
IF they keep their private keys secure. If not, you know nothing.

And even if they do, you only know that they are claiming that it came from their cameras. You have no way of knowing if it actually did. Hence you needing to trust them.

A rogue employee, or secret court order for them to hand over their key to a government agent could have just as well produced a fake image and used their access to the private key to sign it as the camera would've. Plenty of countries have attempted and succeeded in demanding private individuals and companies turn over encryption keys.

Anonymous A (OP) double-posted this 2 years ago, 2 minutes later, 17 hours after the original post[^] [v] #1,277,710

Also, none of what I said would even need to happen. Imagine physically hacking the device and directly manipulating the photosensors. You could just as well inject any image into the camera like this and it would be none the wiser.

And beyond that, one could simply hack the firmware to pull the private key directly. This has also been done in the past to crack DRM. If the stakes were high enough, I'm sure it'd be done again.

(Edited 53 seconds later.)

Anonymous D replied with this 2 years ago, 7 hours later, 1 day after the original post[^] [v] #1,277,816

@previous (A)
Yeah but you can just ask the AI to make one that can't be hacked, retard

Anonymous A (OP) replied with this 2 years ago, 16 minutes later, 1 day after the original post[^] [v] #1,277,818

@previous (D)
umad?

Anonymous D replied with this 2 years ago, 7 hours later, 1 day after the original post[^] [v] #1,277,845

@previous (A)
the AI says I'm not mad, just disappointed

Anonymous H joined in and replied with this 2 years ago, 1 hour later, 1 day after the original post[^] [v] #1,277,859

There’s ways to tell what is made by an AI and people still stupidly fall for footage of Bigfoot or ghosts despite it always being a blurry image in the world of HD cameras.

Anonymous H double-posted this 2 years ago, 1 minute later, 1 day after the original post[^] [v] #1,277,860

@1,277,816 (D)
If the AI can be hacked then how does it make something unhackable, retard.

Anonymous D replied with this 2 years ago, 13 minutes later, 1 day after the original post[^] [v] #1,277,862

@previous (H)
the AI cant be hacked that's why everyone is so worried

Anonymous A (OP) replied with this 2 years ago, 3 hours later, 1 day after the original post[^] [v] #1,277,903

@previous (D)
Just make an AI that hacks the other AI.

Anonymous A (OP) double-posted this 2 years ago, 5 minutes later, 1 day after the original post[^] [v] #1,277,905

@1,277,859 (H)
> There’s ways to tell what is made by an AI
There's something called adversarial training where one network is trained to generate images or video and another is trained to detect if it was generated by an AI or not. The two networks are trained together, simultaneously.

It's an endless cat and mouse game. If you train a network to generate fakes, then there's nothing preventing someone else from coming along and training a larger network with more data to trick yours.

Anonymous D replied with this 2 years ago, 20 minutes later, 1 day after the original post[^] [v] #1,277,909

@previous (A)
there is a little something that can prevent it, called physics

boof replied with this 2 years ago, 2 minutes later, 1 day after the original post[^] [v] #1,277,910

like, what if AI gets curious about what it's like to be blackout drunk or tripping on balls of acid

Anonymous D replied with this 2 years ago, 1 minute later, 1 day after the original post[^] [v] #1,277,911

https://youtu.be/Iw3G80bplTg?si=aJF6KxtvYP9R9b4c

boof replied with this 2 years ago, 2 minutes later, 1 day after the original post[^] [v] #1,277,914

"plug and play muthafucka" was the slogan that Logitech used

Anonymous A (OP) replied with this 2 years ago, 1 hour later, 1 day after the original post[^] [v] #1,277,929

@1,277,909 (D)
Doesn't tell you where the sequence ends though.

Anonymous A (OP) double-posted this 2 years ago, 16 seconds later, 1 day after the original post[^] [v] #1,277,930

@1,277,911 (D)
ytdw

Anonymous D replied with this 2 years ago, 3 hours later, 1 day after the original post[^] [v] #1,277,955

@previous (A)
your're loss

Anonymous I joined in and replied with this 2 years ago, 6 hours later, 2 days after the original post[^] [v] #1,278,045

Anonymous A (OP) replied with this 2 years ago, 36 minutes later, 2 days after the original post[^] [v] #1,278,059

@previous (I)
Lol

Anonymous J joined in and replied with this 2 years ago, 1 day later, 3 days after the original post[^] [v] #1,278,398

:

Please familiarise yourself with the rules and markup syntax before posting.