The truth is this technology is dangerous in the wrong hands. I don't know if that means there needs to be laws created regarding it, but I know that the difficulty in stopping it shouldn't be used as an excuse to not even debate that possibility.
I believe more than 5% are familar with the concept photorealistic computer graphics. Machine learning based face transfer is just a newer iteration that significatnly reduced cost.
Same goes for synthetic voice.
Banning this technology would be more like banning cameras, since they can also be used to create child porn. However, it would be even LESS effective than banning cameras, because we are talking about a software algorithm.
Also, you say this technology is 'dangerous in the wrong hands.' Why is that the case?
I think the only reason is because it has the possibility of decieving people into thinking it is real. The real way to reduce this risk is to either spread the idea that all video is suspect, and you can't assume who you see on video is who it appears to be, or to perfect technology that can instantly detect when this system has been used.
If you are able to easily distinguish between true and false video, isn't the entire danger mitigated?
We also ban computer generated child porn even though there is no victim involved in the production of that. If you use this deepfake technology and stick an 18 year old's face on a porn video it is legal but if you do the same thing with a 17 year old it isn't. The production of those two videos is the same. The only difference is the potential victim if the video became public.
>The real way to reduce this risk is to either spread the idea that all video is suspect, and you can't assume who you see on video is who it appears to be, or to perfect technology that can instantly detect when this system has been used.
That sounds great in theory, but people don't work like that. We are built to believe what we see. We all know that almost every magazine cover has been mercilessly Photoshopped and yet the images on those covers have still been shown to have a strong influence on people's own body image.
For reference, artificially created child porn (e.g. in mangas, or in other ways created) is in a legal grey zone in many areas, including Japan and Germany. It is absolutely illegal in other, e.g. Sweden.
No. The grayness only applies to non-photorealistic works.
Once you start making photorealistic deepfakes or diptychs, where the end result appears to be photographic, the color becomes much more black and white the world over.
But yes, it's a much more complex legal case.
Many places don't do that.
The swap is a "reconstruction" of the face at multiple different angles, taking on the characteristics of the original facial movements, tilt of the head, etc. In a video - way beyond what would be easily possible using PS or other "manual technology".
But today - it's just a face. However, nothing says that an entire person's body couldn't be swapped in the future. Nor that the system couldn't "dream" new scenarios for the background, etc.
Or that it could be done real-time, with full-on voicing in the other person's voice, etc - looking, doing, and saying things they never said - perhaps even in a place they've never been.
In the future, you could easily take someone else's entire person, and have them do jumping jacks in your living room while singing Jingle Bells - and it would likely look like an original video.
The only reason it's not being done now is the amount of processing power that would be needed to do it; gobs more than what most people can afford (along with the time and power needs - just note that for some of those facial swaps it takes hours of processing on a decent system for just a few seconds of video).
I'm not saying this tech is evil or should be banned, however; I think the tech is very neat and can open up some exciting possibilities for future products and uses. To that end, it's exactly like Adobe Photoshop, which is why it should be explored and played with, rather than banned.