Results 1 to 11 of 11
  1. #1
    Senior Member rkspsm's Avatar
    Join Date
    May 2019
    Location
    Delhi, India
    Posts
    987
    Reputation
    1746
    Type
    Propertarian

    The deepfakes gonna be juicy !

    Some improvements in AI algorithms by nvidia :



    If the faces or people generated by that kind of AI of that quality, are presented in a video, I wont be able to distinguish between that and real, unless the AI fucks up in some other way, like hand going into the door, that kind of thing.

    Though the deepfake ecosystem is gonna have a field day with that !!!

    But then I imagine someone creating an orgy video of democrat leaders .... uhh okay, maybe I should get back to work...
    Greedy algorithms always get stuck in a local extrema...

    If you cannot determine global extrema... or if the graph (or space) is infinite, genetic algorithms give better results.

    The function space or hypothesis space is not only infinite, but in most cases, have infinite dimensions.

  2. #2
    Administrator Unboxxed's Avatar
    Join Date
    Mar 2014
    Location
    USA
    Posts
    4,094
    Reputation
    13046
    Type
    enigmatic

    Re: The deepfakes gonna be juicy !

    Yeah, the deepfakes. I'm thinking not about fabricating orgies using familiar faces but about the malicious use of AI to, say, make high-quality deceptive recordings of people's faces whom I trust, saying things they never said, to get me to do things I wouldn't otherwise do.

    I wonder how a person's susceptibility to be "fooled" by this level of computer-generated sophistication may be influenced by one's admiration for it?

    In other words, if I marvel at it, let myself be attracted to it like perhaps the narrator of that video has, am I doomed to be tricked by it, more than if I was indifferent to it? Sounds counter-intuitive doesn't it, because at first thought, one might assume that the more I knew about AI, the more I could recognize it when I saw its misuse. I'd perhaps keep up with it for pleasure but also to know where to expect its applications. Maybe I'd embrace it. Welcome it. Even come to prefer it.

    But, would that dull me to deleterious consequences? Star Trek reference: Dr. Daystrom did not understand that his proudest creation, the M-5 multitronic unit, was not serving man. He was too close to it to see it in any broader perspective.

    It would be going too far in the other direction, though, to prefer to not even know about AI at all, because then for sure I'd be duped by it.

    Should I position myself somewhere in between, perhaps? Will that best protect me for the tricks that are coming?

    That is, is it best to say, meh? To be generally aware of it, giving it just enough notice to keep tabs on it, but not enough that I couldn't steer clear of it. Will that stance help me to ignore it, which may ultimately be the best defense when AI spreads out into our culture?

    In other things, the meh approach I think has served me well. For instance, I've ignored social media and I think I'm much better off for that. I'm aware of the drama going on about it, facebook and the like, so in that sense I keep tabs on it. But, I'm not under its spell. I chose to act like it never was invented. I do that with a lot of different things I hear about, act like they never were invented and then I don't intersect with them. Who needs to absorb all that clutter? Keeps me free from the rip tide that we call modern life.

    But, going back to the AI in this video, it needs a display screen to work its magic. Being tricked requires your assistance and cooperation. I could be ahead of the game here because I belong to that old adage of "Don't believe anything you see and half of what you hear." If I keep myself unimpressed, maybe AI will pass me by.

    No major finish here. Meh.
    The two most important days in your life are the day you were born and the day you find out why. - Mark Twain

    The mass of men lead lives of quiet desperation.
    - Henry David Thoreau

    There are 10 types of people in the world - those who understand binary, and those who don't.

    Suitable for bookmarking: www.fakehatecrimes.org and www.breitbart.com/tag/hate-crime-hoax

  3. #3
    Senior Member rkspsm's Avatar
    Join Date
    May 2019
    Location
    Delhi, India
    Posts
    987
    Reputation
    1746
    Type
    Propertarian

    Re: The deepfakes gonna be juicy !

    Quote Originally Posted by Unboxxed View Post
    In other things, the meh approach I think has served me well. For instance, I've ignored social media and I think I'm much better off for that. I'm aware of the drama going on about it, facebook and the like, so in that sense I keep tabs on it. But, I'm not under its spell. I chose to act like it never was invented. I do that with a lot of different things I hear about, act like they never were invented and then I don't intersect with them. Who needs to absorb all that clutter? Keeps me free from the rip tide that we call modern life.

    But, going back to the AI in this video, it needs a display screen to work its magic. Being tricked requires your assistance and cooperation. I could be ahead of the game here because I belong to that old adage of "Don't believe anything you see and half of what you hear." If I keep myself unimpressed, maybe AI will pass me by.
    From what I know about you, this is probably the best strategy for you. You can simply ignore its existence and dont look at (or care about) things shown on the screen.

    I cannot follow this because, I cannot ignore its existence if I want to keep up with the advances in my profession. Even a hand written program is an AI. Its a rule based AI, which means you describe the entire algorithm. But this approach doesnt scale to harder problems. Face recognition, face generation, etc... you cannot come up with an algorithm to do it.

    What AI does is, it figures out the algorithm. For a very rough analogy think of any math function plotted on a x-y graph, some wobbly wavy-curvy function, which you dont know the formula of. It can be extremely complicated formula and can be almost impossible to figure out. What you can do is... approximate it with a bunch of small straight lines joined together. The shorter the length of lines, the better your approximation is. And once you reach the desired approximation for the problems at hand, you no longer need the exact formula.

    So you see, its an upgrade, from painstakingly figuring out the rules (formulas) to letting the machine approximate it with good enough accuracy and get the job done. I am learning AI these days, from ground up, starting from basic underlying math, and damn its a lot of it. Will likely take a few months if not years to catch up.

    But then to avoid getting into any kind of trap, the best strategy for me is to do a bit opposite from what is suggested in fiction :

    Quote Originally Posted by Unboxxed View Post
    Star Trek reference: Dr. Daystrom did not understand that his proudest creation, the M-5 multitronic unit, was not serving man. He was too close to it to see it in any broader perspective.
    I charge at it, head on. I need to be closer. I need to understand it sooo well, that I can write AIs (or algorithms) to detect AIs. I may not reach this goal, but then I wont be alone doing it. There are researchers who do just this. They devise clever ways to fool the AIs, or to detect deepfakes. I can seek their help, atleast passively by just following their work.

    The old saying about keeping your enemies closer !
    Last edited by rkspsm; July 30, 2021 at 10:22 AM. Reason: spelling
    Greedy algorithms always get stuck in a local extrema...

    If you cannot determine global extrema... or if the graph (or space) is infinite, genetic algorithms give better results.

    The function space or hypothesis space is not only infinite, but in most cases, have infinite dimensions.

  4. #4
    Super Moderator Mr Wombat's Avatar
    Join Date
    Mar 2014
    Posts
    5,070
    Reputation
    23848
    Type
    Neutral

    Re: The deepfakes gonna be juicy !

    What it means is that recorded images are no longer proof of anything. Pictures - even videos - now have the same problem that documents have always had: authenticity. The solution to that is the same as it has always been: a chain of provenance. A video means nothing unless there is a real live human being somewhere declaring "this is an accurate recoding of a thing" - preferably in person right in front of you, but if not, then a documented chain going back to such a person.

  5. #5
    Senior Member rkspsm's Avatar
    Join Date
    May 2019
    Location
    Delhi, India
    Posts
    987
    Reputation
    1746
    Type
    Propertarian

    Re: The deepfakes gonna be juicy !

    Quote Originally Posted by Mr Wombat View Post
    What it means is that recorded images are no longer proof of anything. Pictures - even videos - now have the same problem that documents have always had: authenticity. The solution to that is the same as it has always been: a chain of provenance. A video means nothing unless there is a real live human being somewhere declaring "this is an accurate recoding of a thing" - preferably in person right in front of you, but if not, then a documented chain going back to such a person.
    That, and you can also do some forensics on the video itself, in some cases. Eg, if resolution is quite high, you can zoom onto some object, lets say a tree or a wooden plank or anything with some texture, and then it can be used to either match with some other proof that such texture is indeed realistic to that material, or even more rigorous, can be manually inspected if possible to reach that object.

    And there are many things around us that have textures. Its a kind of weak fingerprinting, not very certain "proof" in itself, but several of them can be presented to increase the certainty that the video is indeed genuine.
    Greedy algorithms always get stuck in a local extrema...

    If you cannot determine global extrema... or if the graph (or space) is infinite, genetic algorithms give better results.

    The function space or hypothesis space is not only infinite, but in most cases, have infinite dimensions.

  6. #6
    Senior Member mgtower's Avatar
    Join Date
    Mar 2019
    Posts
    3,926
    Reputation
    11564
    Type
    Ghosted by law and order.

    Re: The deepfakes gonna be juicy !

    Id' give all the femicrats tree bark texture for skin, pine needles for hair, then set them all on fire!
    Any man that seeks leadership outside himself has a fool for a guide.

  7. #7
    Senior Member
    Join Date
    Jun 2019
    Location
    Wherever you go, there you are.
    Posts
    1,993
    Reputation
    3396
    Type
    Just Me.

    Re: The deepfakes gonna be juicy !

    This has been a concern of mine for a long time. It used to be mostly in the realms of the movie-makers and it took time and a lot of money to make it believable.

    These days though, as technology advances, more and more it will become a resource available to the (slightly more than) average Joe.

    It brings a whole new meaning to the phrase "I can't believe my eyes."

  8. #8
    Senior Member rkspsm's Avatar
    Join Date
    May 2019
    Location
    Delhi, India
    Posts
    987
    Reputation
    1746
    Type
    Propertarian

    Re: The deepfakes gonna be juicy !

    Quote Originally Posted by Jackoff View Post
    This has been a concern of mine for a long time. It used to be mostly in the realms of the movie-makers and it took time and a lot of money to make it believable.

    These days though, as technology advances, more and more it will become a resource available to the (slightly more than) average Joe.

    It brings a whole new meaning to the phrase "I can't believe my eyes."
    I have a concern, well not much of a concern but curiosity about a bit darker usage of this on average joe level. He (or she) will be able to bring to life (on the screen, or maybe even in VR) any weird fantasy of anybody else in family, friends, neighbors, colleagues, etc...

    Superficially it may not sound so bad, but it has potential to have significant effects on psychology of people, especially those who are not capable to properly guard their emotional state and are easily influenced (which is, most of the people). The fantasy can involve any kind of act.

    Initially it will be limited by the options available on the apps, but eventually driven by market competitions, apps will get creative, and more and more options will open up. And that is when it will get really crazy really fast...
    Greedy algorithms always get stuck in a local extrema...

    If you cannot determine global extrema... or if the graph (or space) is infinite, genetic algorithms give better results.

    The function space or hypothesis space is not only infinite, but in most cases, have infinite dimensions.

  9. #9
    Senior Member
    Join Date
    Jun 2019
    Location
    Wherever you go, there you are.
    Posts
    1,993
    Reputation
    3396
    Type
    Just Me.

    Re: The deepfakes gonna be juicy !

    Quote Originally Posted by rkspsm View Post
    I have a concern, well not much of a concern but curiosity about a bit darker usage of this on average joe level. He (or she) will be able to bring to life (on the screen, or maybe even in VR) any weird fantasy of anybody else in family, friends, neighbors, colleagues, etc...

    Superficially it may not sound so bad, but it has potential to have significant effects on psychology of people, especially those who are not capable to properly guard their emotional state and are easily influenced (which is, most of the people). The fantasy can involve any kind of act.

    Initially it will be limited by the options available on the apps, but eventually driven by market competitions, apps will get creative, and more and more options will open up. And that is when it will get really crazy really fast...
    Oh I’m thinking MUCH worse than that, similar to what Unboxxed said about deceptive recordings.

    And forget apps (if you’re referring to those shitty things you can download to your phone as opposed to major computer applications). Full blown movie editing suites are available even today, still limited compared to a movie studio but hell you can do a lot with them. Think NERO on steroids.

    Political enemies, including but not limited to electoral politics, having videos made superimposing their rivals doing nefarious things or even just saying things that they would never say.

    Or, imagine a scenario where two employees are vying for the same promotional post. Throw something up on a social media site or youtube to knock your opponent out of contention.

    Or, what of fake rape allegations? O.K. the vid may be found out to be fake and so not allowed as court evidence, but if the jury pool is already tainted?

    It wouldn’t really matter if the video was eventually proved fake, the damage to public and/or employer perception would be done.

    Visual evidence, even when faked, is VERY persuasive.

    On the plus side though, you’ll get many intentionally humorous vids of well-known people doing REALLY strange shit.

  10. #10
    Senior Member rkspsm's Avatar
    Join Date
    May 2019
    Location
    Delhi, India
    Posts
    987
    Reputation
    1746
    Type
    Propertarian

    Re: The deepfakes gonna be juicy !

    Quote Originally Posted by Jackoff View Post
    And forget apps (if you’re referring to those shitty things you can download to your phone as opposed to major computer applications). Full blown movie editing suites are available even today, still limited compared to a movie studio but hell you can do a lot with them. Think NERO on steroids.

    Political enemies, including but not limited to electoral politics, having videos made superimposing their rivals doing nefarious things or even just saying things that they would never say.
    What you say can happen, but I still think the impact on personal scale can be more problematic than on celebrity scale. Reason why I think that is, morphing of celebrity images happen even today. And the actual nerds who are into that stuff, and skilled with that, will start with celebrity way early than an average dude getting his hands on an easy to use app. Even today you can find deepfakes of celebrities fucking like rabbits with good enough quality. And if that is not enough, many celebrities have history which can be found by simple google search. Eg: Melania Trump. Google her if you dont know...

    With easy to use app however, an average teen might wanna try hooking up his friend X with friend Y's mother, or sister, or aunt. And this can go two ways. One is when even this is gets so overused that nobody will care eventually. OR... something else entirely which is unexpected. This has never been done before. There are entire cultures based around maintaining atleast an image of sexual modesty around family. Nobody can say how this will impact them. Their kids already have the phones, and once the app is available, everyone will have them within few days.
    Greedy algorithms always get stuck in a local extrema...

    If you cannot determine global extrema... or if the graph (or space) is infinite, genetic algorithms give better results.

    The function space or hypothesis space is not only infinite, but in most cases, have infinite dimensions.

  11. #11
    Senior Member
    Join Date
    Jun 2019
    Location
    Wherever you go, there you are.
    Posts
    1,993
    Reputation
    3396
    Type
    Just Me.

    Re: The deepfakes gonna be juicy !

    You may be right.

    But when it comes to celebrity figures, especially politicians, the affects could decide the fate of countries and their whole populations with just a single, well thought out vid released at exactly the right time.

    That being said, if it happens widespread in communities on an individual level this could also affect people’s mentalities.

    It’s difficult to know which would be more detrimental to society in general, but it seems we may soon find out.


Similar Threads

  1. Replies: 4
    Last Post: March 6, 2021, 2:42 PM
  2. Replies: 0
    Last Post: March 6, 2021, 1:52 AM
  3. Woman's day? Im gonna puke.........
    By MGTOWFOREVER in forum Lounge
    Replies: 10
    Last Post: March 9, 2018, 2:29 PM
  4. Replies: 11
    Last Post: February 11, 2016, 2:57 AM
  5. Replies: 20
    Last Post: February 11, 2015, 11:47 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •