lainchan / Ω
New voice algorithms
SHODAN · 2019-05-16 20:49
https://www.youtube.com/watch?v=DWK_iYBl8cA
https://www.youtube.com/watch?v=cfhFuqF0tEs
Man, we are fuarrrked. Especially our elderly relatives, phone scams will reach a whole different level.
Soon anyone will be able to train an algorithm with your relatives' voices, and by using only a few minutes worth of audio (either from an infected smartphone or some publicly available video they might've shared online) call them with a near-perfect copy of anyone's voice. Or make a skype/facetime call with a deepfaked face of anyone.
Neuromancer · 2019-05-16 21:59
>shodan
Nice.
It's survival of the fittest really, for every bit of technology or social change it is the same. People predicted disaster, but only a part of society that can not keep up with the times fall victim. This one just seems more interesting because we are entering the world of science fiction in a much more tangible way finally.
Drifter · 2019-05-17 01:40
I think I mentioned this in a previous thread, but what worries me the most about this is how it can be used to fake and dismiss evidence that someone said something damning.
If you had access to this software, a vendetta against someone, and preferably more social influence than them, then you could make a fake audio file of them saying something horrible or embarrassing to damage their reputation. Or even worse, make a file of them confessing to a crime they're accused of. And conversely, if you actually did say something damning and you're socially skilled enough, you can convince people that this recording of you was actually faked by software. Of course, that last one might be difficult since people have been making a habit of archiving things to counter this very thing, but it's still possible.
Lady 3Jane.. · 2019-05-17 03:49
>>17257
>e. Of course, that last one might be difficult since people have been making a habit of archiving
how would that help? there would still be doubt.
after all those articles about using software for revenge porn, people were worried we won't be able trust to videos anymore... we aren't at that point yet, but these technologies will challenge how we trust evidence
Emmanuel · 2019-05-18 14:06
>>17253
I heard about people using this to gain access to bank accounts a few years back. Why did it make the news again?>
Unspecified · 2019-05-22 12:22
>>17254
I recently watched a TED talk about deepfakes, and the speaker said that they'd developed a browser addon which would flag potentially deepfaked pictures and videos. I don't know how well that particular addon works, but I think this might be the the only successful approach to combating it.
A highly trained algorithm that could discern fake audio, video and images could be implemented across YouTube/Twitter/Skype/etc. and notify users that whatever they're watching/listening might be faked. If using deepfaked audio to make phone calls is made illegal, the same algorithm could be installed to monitor phone lines and terminate phone calls when it detects faked audio.
Would making an algorithm like this open source make it easier to bypass, though?
Cypher · 2019-05-22 13:02
>>17327
>Would making an algorithm like this open source make it easier to bypass, though?
far from an expert, but i guess it depends on how it'd work? if it analyzed audio and found unnatural elements that are caused by generation, probably not that much... because you wouldn't change it just to deviate from what your anti-deepfake algorithm looks for, but the whole process how you generate it. i mean, i think i read about something like that for pictures already and that's something they mentioned i think... so i am basing it off that.
anyone who actually knows anything about this?
Wintermute · 2019-05-23 02:17
How long before this is used to generate scandal around political figures & for espionage? Yesterday?
This + advanced, government secret deepfakes mean video and audio are soon going to be as "factual" as plain written text. So on the one hand, hyperreal dystopia. But on the other, it can't be much worse than the way things were a hundred years ago with yellow journalism (once people adapt and learn to disbelieve everything).
Even without this specific technology, we've had tons of fakes throughout history and recent events. Gulf of Tonkin, Iraq Nukes, etc. And then in the Syrian war, all the staged ISIS videos, fake chemical attack videos, etc. with complete sets and greenscreens and actors.
In order to discern reality from the fakes, you need to know history, have a real theory of politics and society, and compare everything to what you can actually see in front of you IRL. I posit that dialectical and historical materialism is the only weapon you can arm yourself with to combat this spectacle.
Roy Batty · 2019-05-23 02:21
Also, this is gonna be great for Vent Harassment type stuff.
https://www.youtube.com/watch?v=2TkifdDCwe8
Come to think of it, these Vent Harassments are like COINTELPRO 101.
Guthrie · 2019-05-23 03:56
>>17340
This is kind of what I was saying in the first reply. If deepfakes become part of our lives in politics, wars etc or even in daily life with scammers etc it will just change the paradigm on what people believe and what they don't.
Yet Another Lainon · 2019-05-23 04:03
>>17347
It would be great, forcing face to face communication again once voice becomes too much of an inconvenience due to scammers, and helping finish this loneliness epidemic.
The sad part is exactly that, only scam will change our ways as it messes up directly with our convenience. Politics/wars/etc? Please, on current time whoever makes the most aggresive fanfiction is gathering the attention, truth abandoned ages ago.
Crash Override · 2019-05-23 22:02
New face algorithm by Samsung. It feels like the development of this soykaf is escalating exponentially:
https://www.youtube.com/watch?v=p1b5aiTrGzY
Y.T · 2019-05-23 22:52
A single image as source is just crazy stuff, wonder what is the potential beyond producing fake content.
Yu-ichi · 2019-05-24 03:08
>>17356
Governments of the world will use this and there will be nothing anyone can say about it.
Koichiro · 2019-06-08 13:28
>>17340
The problem with your solution is that there is very little we can positively verify in meatspace these days without going to absurd lengths. Dismissing any evidence that contradicts one's worldview as fake unless it can be personally verified is pretty much exactly what those who are trying to break us into tribes want.
Ivan Petrov · 2019-06-08 22:37
>>17348
>It would be great, forcing face to face communication again once voice becomes too much of an inconvenience due to scammers, and helping finish this loneliness epidemic.
my first thought too
Lady Jessica · 2019-06-09 09:59
>>17367
So will private companies etc etc.
What a time to be alive.
John Rain · 2019-06-09 21:06
>>17597
>>17348
...or it will force people into deeper and deeper isolation and paranoia. If that level of deception becomes commonplace, people will believe everything else is a trick too.
Nameless · 2019-06-10 15:53
New paper which shows insertion and deletion of sentences:
https://www.youtube.com/watch?v=0ybLCfVeFL4