Hacker News Reader

Comment by rndmize | Hacker News Reader

Contrary to the article's stance that this would be used to create fake revenge porn or celebrity sex tapes that are taken as real, I'd expect this to devalue the real thing and give space for people to claim anything with their face in it is fake.

View on HN

We've already seen this with images and Photoshop. Society and their heuristics of belief will adjust as these new capabilities become widespread.

What's more troubling is that as media becomes falsifiable, solid evidence of...well, anything, becomes hard to have. The ultimate loser there is the truth, sadly.

Indeed. The worst use of this will be forged political speeches. What happens when you can fake any video of anyone saying anything and then feed it to the social media outrage machine?

As the parent noted, society will adjust expectations about how to determine facts. When radio first came along, some people were fooled that Orson Welles's War of the Worlds [1] was actually happening, because it was in the style of a news broadcast. We learn and adjust how to process data as systems change around us.

https://www.smithsonianmag.com/history/infamous-war-worlds-r...

Yes, but that in no way implies that the method/means society adjusts it's behaviours/expectations will be desirable.

One could even hypothesise that social media bubbles are the first step precisely in that direction.

You can already fake a political speech if you've got the resources. A look-alike and a good make-up artist can do a lot. The danger isn't a really good fake that can be discussed, but a high volume of OK fakes.

Its really not so different than it was prior to the advent of radio. By the time news reached you about what some politician said it had gone through multiple people and could be altered in any number of ways. Sometimes intentionally and sometimes through misunderstanding.

I've suspected for a long time that much of what we know about history from written accounts is false.

This is a basic assumption historians take. They assume sources have biases and agendas, and try to figure out what is true. It isn't a perfect system, but no historian takes a source at face value.

The question that concerns me more than Terminator- and Matrix-like scenarios is why we allow "the social media outrage machine" to exist in the first place. It is nothing more than a bunch of mostly uneducated opinions that create erratic views around the world of their M-sized subscriber sets.

Maybe this new tech will force people to bet their credibility on what they say and spread.

The truth can benefit greatly if the standard for fact checking changes. Media is not a great friend to truth, especially not when it conflict with either revenue or political direction.

Just take the massive global interest in fact checking sites after the 2016 US election. It seems as a strong trend in competing with media for carrying the banner for truth, and the best part is that many such sites seems to have a non-partisan policy in stark contrast to most media out there. The loser don't seem to be the truth, but rather those who wish to use a few seconds of video with no context to push a point.

Trolls seem to be doing just fine. It's also cool to be a fact checker but if you think the fakers have lost then you're naive, they're stronger than ever.

It's horrible, people will have to have reputations, be careful about what they say, and consider who they trust again.

It wasn't the good old days or anything, but that was how it worked for a while. There will be value in being a trustworthy source of information again.

> The ultimate loser there is the truth, sadly.

No, it will force the development of AI to detect whether things are fake or not.

We'll soon be in an arms race as better-and-better AIs alternately try to fool or not be fooled.

The people who are able to understand and wield these technologies will have incredible amounts of power over those who are not.

This is one route to the singularity - to deal with the consequences of ML and AI advancements, you need to invest even more. This ever-increasing arms race sucks vast amounts of human and financial capital into the 'tech sphere'.

I think we're already seeing a lot of cases where even AI to detect fakes will be irrelevant if the viewer wants to believe or ignore the alternative.

That ignores the role of actual social networks and social proof in the construction of consensus reality. "Our" AI says it's real, and "your" AI says it's fake, and one can debate the margins - just as one debated whether it was metaphysically true that Clinton didn't have "sexual relations" with Lewinsky, or that Bush "lied" about Iraq.

Another option is that we will develop both cynicism about claims and care less about nudes/porn

a lot of this stuff is already generated using generative adversairal networks, so i think the AI to fool is just as good as the one thats trying not to be fooled by design.

I started pursuing a PhD application on this topic, speaking to a couple of potential supervisors, before losing interest.

Ideas: https://github.com/pjlsergeant/multimedia-trust-and-certific...

You could make a market out of tamper proof blockchained cameras?

i often wonder if there is a way for people to put information on the blockchain so it -is- verifiable (you can post a public key to demonstrate that the information is from you). so that all public information is only official if its on the blockchain. soon enough people learn that they can't trust anything that is not on blockchain.

of course, talk is cheap. i dont have the chops to actually spec this out.

did you know you can make digital signatures without a blockchain

immutability/permanent record

What do you mean? A change to digitally signed data can be detected immediately without any blockchain.

Yeah and then someone puts something fake on the blockchain. Now what?

then it wouldnt be signed by them

"Oh that porn video of me isn't signed. My signed porn videos are over there"

How are you not seeing the obvious issues with your idea? People are not going to put compromising photos and videos on this blockchain because they are compromising. And yet some compromising photos and videos taken by other people are real. The fact that it's "on the blockchain" means basically nothing in this case.

Other users who were discussing "hardware signing" are closer to a working system. This is not a problem for a blockchain to solve.

This is where hardware signing comes into play.

Did you mean "less falsifiable"?

I'd like to believe that, but I suspect that more likely would be that people believe the videos of their friends are fake and believe the videos of their enemies are real, pushing us further down the easy slope of polarization by giving it even more fodder to work with.

Pretty much what happened with UFO photos, ghost photos, etc.

Why not both?

Reply on news.ycombinator.com