New Algorithm Converts Portraits Into Motion Videos

in #technology5 years ago

1.jpg

A few days ago I had written a post about this AI system that could synthesise sound clips from text and make it sound like anybody its creators wanted and that the researchers had used the voice of Joe Rogan as an example.

Although impressive, I found it really creepy that you could essentially make anybody say anything which could have huge repercussions. Well, if that was not enough to scare you, there has been a new development that takes things a step further.

Egor Zakharov has created an algorithm that can covert still portraits of people into short videos. If you're wondering where you might have seen this before, it's probably in the Harry Potter movies where a photograph of a person, acts as a video in newspapers and where paintings come alive.

You may already know about deepfakes where machine learning algorithms can be used to analyse a video of a person and then their face can be overlayed onto other person's face in another video. It's essentially a face swap in very simple terms, but this algorithm goes even beyond that.

Right Out of Harry Potter

By feeding still photos of a person, the machine learning algorithm is able to analyse the different facial features and predict what they would look like in various angles, something that is critical if a convincing video is to be made.

The more photos are fed into the system, the better the results are. What is amazing is that the algorithm works even if it is fed just one photo although the result can easily be seen as something generated by a computer.

You can check out in the video above how great the system works with even just three photos of someone. But what you will find really amusing is that this algo can be used to make videos of paintings as well. It was able to make Mona Lisa talk!

Again, as with the previous AI system, the same problems exist here also. We are quickly moving to a time where the authenticity of nothing could be sure about. Not videos, not photos and not even sound clips. I wonder how the legal system, security systems or even society as a whole will adjust to this new reality but I hope enough timely regulations are made so that the common folk don't end up paying the price for something they didn't do.

Sort:  

To listen to the audio version of this article click on the play image.

Brought to you by @tts. If you find it useful please consider upvoting this reply.

Thanks!

Oh my gosh again this is scary AI tech that needs to be shared a lot more and talked about a lot more. The progress of this technology needs to slow down so lawmakers, and other powerful individuals and communites can regulate them and make truth a prioity. This stuff is real and coming faster than most folks may think. Great job posting about this very interesting but again very scary stuff. I have hope we can find a way to keep it safe.

Yeah, most people don't even know that these things exist already and there needs to be a lot more sharing of knowledge and info about this stuff. Only then can we take timely action otherwise things will get out of hand.

You got a 41.77% upvote from @ocdb courtesy of @sauravrungta! :)

@ocdb is a non-profit bidbot for whitelisted Steemians, current max bid is 30 SBD and the equivalent amount in STEEM.
Check our website https://thegoodwhales.io/ for the whitelist, queue and delegation info. Join our Discord channel for more information.

If you like what @ocd does, consider voting for ocd-witness through SteemConnect or on the Steemit Witnesses page. :)

Coin Marketplace

STEEM 0.19
TRX 0.16
JST 0.030
BTC 80986.05
ETH 3142.29
USDT 1.00
SBD 2.74