Underestimating the mind-warping potential of fake video • Hacker News Board

Re: Underestimating the mind-warping potential of fake video

#101

Post by baq » Mon, Apr 23, 2018, 6:37 PM UTC

lolc wrote: post #72

rossdavidh wrote: post #42

Earlier quoted context omitted.

My thought is that it is much like what to do about any new technological threat, for example too many calories or too much information or too many distractions while driving. Bad things happen, and after about a generation coping mechanisms get developed, and some people adopt them more successfully than others. One is reminded of how populations adapt to new diseases. One valid question, though, is whether or not t…

We're able to distinguish entertainment "fake" video from "real" video in our minds. The separation is not clear-cut, and fictional movies affect our thinking about real events. So I understand the concern. But faking video is not a new invention. It's happened before and in a way I welcome the broad availability of faking-tools to inoculate against the fakes that have been with us since video existed. Now everybody…

Now everybody will be able to create convincing fakes. That's not continuous progress, that's a phase change.

Re: Underestimating the mind-warping potential of fake video

#102

Post by zuminator » Mon, Apr 23, 2018, 6:38 PM UTC

lolc wrote: post #25

There will be a plethora of "Deepfake" apps for your pocket computer tomorrow. People will learn quickly not to trust video. They will have no choice because they don't want to fall for pranks. Humanity will stand through this.

I think you overestimate people. We DO want to be pranked. We want to believe powdered rhino horn will cure impotence, we want to believe the Moon landings were faked, we want to believe in that Florida property we purchased sight unseen, we want to believe our grandparents are in the afterlife. And we will damn sure want to believe in videos of despised evil politicians biting the heads off fetuses they ripped out of the womb with their bare hands.

In some respects lies are easier to believe than the truth because they don't rely on facts at all.

Re: Underestimating the mind-warping potential of fake video

#103

Post by lifeisstillgood » Mon, Apr 23, 2018, 6:42 PM UTC

Is there not a chain of trust here that can be enforced with fairly simple technology. And then the trust in the technology is something like a "kite mark" - a global standard of handling video content.

Let's take a simple example - I am a journalist who videos Obama talking. My raw video footage is tagged in each frame with a hash of the pixels in the frame. then I process the video, making linear cuts etc.

If i publish the whole lot as a package, anyone else can come along and rerun the processing and check the hashes match up.

This should be decentralisable and scalable. and we just have to get used to clicking the kitemark to check if anyone has verified the process or even run it ourselves.

one could imagine a cottage industry of verifiers

Re: Underestimating the mind-warping potential of fake video

#104

Post by CM30 » Mon, Apr 23, 2018, 6:45 PM UTC

GCU-Empiricist wrote: post #19

beamatronic wrote: post #12

Earlier quoted context omitted.

Perhaps the model for “trusted news” would follow other established certificate signing models

So what would the trust criteria be? There is a lot of credible disagreement about trustworthy media sources today after all.

Especially given that in the modern media landscape, standards for what's considered trustworthy have changed significantly. Think about it, before it was usually 'did this journalism thing professionally without having a conflict of interest', and that's roughly how the likes of Wikipedia define reliable sources.

But now that's not the case, and in many cases, it's amateurs who more trustworthy than the so called professionals are. That's true in science reporting (where it's often academics writing blogs and stuff outside of their university employment), it's true in technology (where many reliable sources are run by hobbyists) and it's true of the gaming and entertainment media worlds where fan sites, blogs and YouTube channels are often far more reliable than large media organisations are.

A good solution here would basically need to be able to figure out that Science Blogs is more reliable than say the Daily Telegraph when it comes to science reporting, or that Serebii.net is more likely to be right about Pokemon than say the Guardian or BBC is.

Re: Underestimating the mind-warping potential of fake video

#105

Post by pjc50 » Mon, Apr 23, 2018, 6:46 PM UTC

gameswithgo wrote: post #48

uptown wrote: post #41

There's a relevant and insightful thread on Twitter posted by David Roberts here: https://twitter.com/drvox/status/988144274720964608 His thread references the "debate" over climate change, and his point is that the debate most people participate in isn't truly a debate over facts, but rather a debate over trust. That matters of complexity are too distant from any one individual to have direct exposure-to, so in orde…

>That matters of complexity are too distant from any one individual to have direct exposure to In the case of climate change it is actually possible to self-inform yourself about some of the key aspects of climate change. Raw data is freely available (noaa and others), and some simple scripts can answer questions like "are the claimed warming trends real or artifacts of bias due to adjustments or other factors". This…

> some simple scripts can answer questions like "are the claimed warming trends real or artifacts of bias due to adjustments or other factors".

This requires at least a graduate-level stats education to do "properly"!

Re: Underestimating the mind-warping potential of fake video

#106

Post by jpfed » Mon, Apr 23, 2018, 6:48 PM UTC

snowwrestler wrote: post #45

It seems like fake video can be handled the same way we handle fake text. We're just not used to doing that yet. Anyone can write an article and sign it "Barack Obama," and post it online somewhere. Not many folks will fall for it. Most people understand how easy it is just write someone else's name down, so they look for provenance to help decide what is real. An article by "Barack Obama" on the Wall Street Journal…

Here's "Barack Obama" giving a thumbs-up to repealing net neutrality: https://www.fcc.gov/ecfs/filing/1051157755251

Re: Underestimating the mind-warping potential of fake video

#107

Post by godelski » Mon, Apr 23, 2018, 6:53 PM UTC

uptown wrote: post #51

gameswithgo wrote: post #48

Earlier quoted context omitted.

>That matters of complexity are too distant from any one individual to have direct exposure to In the case of climate change it is actually possible to self-inform yourself about some of the key aspects of climate change. Raw data is freely available (noaa and others), and some simple scripts can answer questions like "are the claimed warming trends real or artifacts of bias due to adjustments or other factors". This…

> "Raw data is freely available (noaa and others)" But still - this is a matter of whether one chooses to trust NOAA and others. His point isn't really even about climate change. It's about debating the merits of any position, and what sources of information you trust to reinforce your position.

> this is a matter of whether one chooses to trust NOAA and others

I should mention that this is fairly easy to verify, at least with part of the data. For land temperatures one of the big sources they use is meteorological data gathered at airports (METAR). You know, the same data that pilots use to calculate take off and landing distances, altimeter settings, etc. So a concerned citizen could compare this these data sets and verify part of the NOAA data. Faking this data, at the airports, would cause disasters, unless you had this major conspiracy where you fixed the pilots instruments or every pilot was aware of this faking and used the real numbers. But both seem really silly since GA pilots need this information too, and frequently do hand calculations.

I certainty think the verify part is essential in phrase "trust, but verify". I just want it to be clear that one can trivially verify one of the large components of the data set.

[1] SFO airport data: http://www.wx-now.com/weather/wxcurrent?icao=KSFO

Re: Underestimating the mind-warping potential of fake video

#108

Post by jpfed » Mon, Apr 23, 2018, 6:55 PM UTC

lifeisstillgood wrote: post #103

Is there not a chain of trust here that can be enforced with fairly simple technology. And then the trust in the technology is something like a "kite mark" - a global standard of handling video content. Let's take a simple example - I am a journalist who videos Obama talking. My raw video footage is tagged in each frame with a hash of the pixels in the frame. then I process the video, making linear cuts etc. If i pub…

I want to believe that something like this would work. Maybe I'm missing something, but isn't it the case that whatever signing could be performed on data coming from CCDs could also be performed on data coming from a GAN?

Re: Underestimating the mind-warping potential of fake video

#109

Post by CM30 » Mon, Apr 23, 2018, 7:00 PM UTC

acjohnson55 wrote: post #69

I think it's important to note that the truth has never been widely available, to the point where it might be useless to think of truth as being a real concept to begin with. One could say that for a blip of time, we lived in a world where a large percentage of human were pushed largely accurate (if incomplete) information by mass media, allowing for something resembling consensual reality and informed democracy. We…

Similarly, it's also worth noting that the means of verifying a story or claim weren't widely available for a long time either. Seriously, go back to before photographs became a thing in 1839 or so. How did you verify any claims at all?

You usually just took people's word for it, or assumed their writings/paintings/memories were accurate. Unless you had the means to actively verify a claim yourself and the claim was about something you still could verify (aka the people involved weren't dead and the organisations and places there still existed), you never really could be sure.

And the same sort of setup goes for other forms of evidence too. Audio evidence, video evidence, even things vital to crime solving now like fingerprinting and DNA testing are all new in the greater scheme of things. For most of human civilisation, being at least somewhat unsure of what's true and what isn't has been the norm, and the time periods where forms of evidence were easy to obtain and hard to fake few and far between.

So yeah, it's definitely like a reset back to the olden days, and the challenges are certainly more about how to deal with informed democracy becoming an illusion again.

Re: Underestimating the mind-warping potential of fake video

#110

Post by Bartweiss » Mon, Apr 23, 2018, 7:03 PM UTC

uptown wrote: post #51

gameswithgo wrote: post #48

Earlier quoted context omitted.

>That matters of complexity are too distant from any one individual to have direct exposure to In the case of climate change it is actually possible to self-inform yourself about some of the key aspects of climate change. Raw data is freely available (noaa and others), and some simple scripts can answer questions like "are the claimed warming trends real or artifacts of bias due to adjustments or other factors". This…

> "Raw data is freely available (noaa and others)" But still - this is a matter of whether one chooses to trust NOAA and others. His point isn't really even about climate change. It's about debating the merits of any position, and what sources of information you trust to reinforce your position.

> whether one chooses to trust NOAA and others

One of the biggest objections from (science-literate) people who reject climate change is precisely this - that 'methodology corrections' applied to NOAA data create a warming trend in underlying data that lacks it. The trust question applies at every level down to "the guy who goes out and looks at the thermometer".

Appeals to "you can run the data yourself" honestly feel like they're missing the point. Pretty much everything we know has some element of faith to it - some of it we can't verify directly, and almost all of it we won't verify, because that task is impossibly large.

Hell, I can't prove that China exists without relying on trust in others. Which is not exactly an academic question when you meet people who say things like "North Korea isn't repressive, that's just a lie our capitalist government tells us". It's not a hopeless case, you can try to push claims to the point where falseness would require a completely implausible conspiracy, but it's worth remembering that "just look at the data!" usually amounts to a call of "just trust these voices!"