Robin Williams’ daughter Zelda says AI recreations of her dad are ‘personally disturbing’::Robin Williams’ daughter Zelda says AI recreations of her dad are ‘personally disturbing’: ‘The worst bits of everything this industry is’

  • Blapoo@lemmy.ml
    link
    fedilink
    English
    arrow-up
    141
    arrow-down
    3
    ·
    1 year ago

    Disturbing is an understatement. I’d call them repulsive. Relatives should be the only ones with this power, if at all.

    Sure as shit not corporations. Fuck.

    • whatwhatwhatwhat@lemmy.world
      link
      fedilink
      English
      arrow-up
      47
      arrow-down
      2
      ·
      1 year ago

      Agreed, we desperately need regulations on who has the right to reproduce another person’s image/voice/likeness. I know that there will always be people on the internet who do it anyway, but international copyright laws still mostly work in spite of that, so I imagine that regulations on this type of AI would mostly work as well.

      We’re really in the Wild West of machine learning right now. It’s beautiful and terrifying all at the same time.

        • TwilightVulpine@lemmy.world
          link
          fedilink
          English
          arrow-up
          18
          ·
          1 year ago

          Copyright IS too strong, but paradoxically artists’ rights are too weak. Everything is aimed to boost the profits of media companies, but not protect the people who make them. Now they are under threat of being replaced by AI trained on their own works, no less. Is it really worth it to defend AI if we end up with less novel human works because of it?

          • lloram239@feddit.de
            link
            fedilink
            English
            arrow-up
            1
            arrow-down
            6
            ·
            edit-2
            1 year ago

            Now they are under threat of being replaced by AI trained on their own works, no less.

            And they themselves trained on the work of other artists too. It’s just the circle of life. AI just happens to be better at learning than humans.

            Is it really worth it to defend AI if we end up with less novel human works because of it?

            AI doesn’t need defending, it fill steamroll us all just by itself. We don’t really have a choice in this.

            • TwilightVulpine@lemmy.world
              link
              fedilink
              English
              arrow-up
              6
              ·
              1 year ago

              The “circle of life” except that it kills the artists’ careers rather than creating new ones. Even fledgling ones might find that there’s no opportunity for them because AIs are already gearing to take entry-level jobs. However efficient AI may be at replicating the work of artists, the same could be said of a photocopier, and we laws to define how those get to be used so that they don’t undermine creators.

              I get that AI output is not identical and its output doesn’t go foul under existing laws, but the principles behind them are still important. Not only Culture but even AI itself will be lesser for it if human artists are not protected, because art AIs quickly degrade when AI art is fed back into it en masse.

              Don’t forget that the kind of AI we have doesn’t do anything by itself. We don’t have sentient machines, we have very elaborate auto-complete systems. It’s not AI that is steamrolling artists, it’s companies seeking to replace artists with AIs trained on their works that are threatening them. That can’t be allowed.

              • lloram239@feddit.de
                link
                fedilink
                English
                arrow-up
                2
                arrow-down
                6
                ·
                1 year ago

                The “circle of life” except that it kills the artists’ careers rather than creating new ones.

                It will kill all the ones that are stuck on old technology. Those that can make the best use of AI will prevail, at least for a little while, until AI replaces the whole media distribution chain and we’ll just have our own personal Holodeck with whatever content we want, generated on demand.

                However efficient AI may be at replicating the work of artists, the same could be said of a photocopier

                It’s not copying existing works and never did. Even if you explicitly instruct it to copy something existing, it will create its own original spin on the topic. It’s really no different than any artist working on commission.

                Don’t forget that the kind of AI we have doesn’t do anything by itself.

                You are free to ignore the reality of it, but that’s simply not the case. AI systems are getting filled with essentially all of human knowledge and they can remix it freely to create something new. This is the kind of stuff AI can produce just by itself, within seconds, the idea is from AI and so is the actual image. Sentience is not necessary for creativity.

                It’s not AI that is steamrolling artists, it’s companies seeking to replace artists with AIs trained on their works that are threatening them.

                When the artists are that easy to replace, their work can’t have been all that meaningful to begin with.

            • wanderingmagus@lemm.ee
              link
              fedilink
              English
              arrow-up
              1
              ·
              1 year ago

              Steamroll us? Not if I have anything to say about it. I look forward to setting condition 1SQ for strategic launch of thermonuclear weapons. Hooyah navy. If this is to be our end, then let it be SUCH an end, so as to be worthy of remembrance. I would rather this entire planet and all things upon it burn in radioactive fire than be sacrificed on the altar of technology.

      • lloram239@feddit.de
        link
        fedilink
        English
        arrow-up
        8
        ·
        1 year ago

        but international copyright laws still mostly work in spite of that, so I imagine that regulations on this type of AI would mostly work as well.

        The thing is, people still don’t grasp the ease with which this will be possible and to a large degree already. This doesn’t need hours of training anymore, you can clone voices with three seconds of audio and faces from a single image. Simple images can be clicked together in seconds with zero effort. Give it a few more years and you video can be created with equal ease.

        You can regulate commercial use of somebodies likeness, which it largely already is, but people doing it for fun is unstoppable. This stuff is here today and it will get a whole lot more powerful going forward.

        • vidarh@lemmy.world
          link
          fedilink
          English
          arrow-up
          2
          ·
          1 year ago

          Just a few years back, Vernor Vinge’s scifi novels still seemed reasonably futuristic in dealing with the issue of fakes well by including several bits where the resolution of imagery was a factor in being able to analyze with sufficient certainty that you were talking to the right person, and now that notion already seems dated, and certainly not enough for a setting far into the future.

          (at least they don’t still seem as dated as Johnny Mnemonic’s plot of erasing a chunk of your memories to transport an amount of data that would be easier and less painful to fit in your head by stuffing a microsd card up your nose)

      • banneryear1868@lemmy.world
        link
        fedilink
        English
        arrow-up
        5
        ·
        1 year ago

        IMO people doing it on their own for fun/expression is different than corporations doing it for profit, and there’s no real way to stop that. I think if famous AI constructs become part of big media productions, it will come with a constructed moral justification for it. The system will basically internalize and commodify the repulsion to itself exploiting the likeness of dead (or alive) actors. This could be media that blurs the line and proports to ask “deep questions” about exploiting people, while exploiting people as a sort of intentional irony. Or it will be more like a moral appeal to sentimentality, “in honor of their legacy we are exploiting their image, some proceeds will support causes they cared about, we are doing this to spread awareness, the issue they are representing are too important, they would have loved this project, we’ve worked closely with their estate.” Eventually there’s going to be a film like this, complete with teary-eyed behind-the-scenes interviews about how emotional it was to reproduce the likeness of the actor and what an honor it was. As soon as the moral justification can be made and the actor’s image can be constructed just well enough.

      • _number8_@lemmy.world
        link
        fedilink
        English
        arrow-up
        5
        ·
        1 year ago

        yeah i don’t think it should be legislated against, especially for private use [people will always work around it anyway], but using it for profit is really, viscerally wrong

      • assassin_aragorn@lemmy.world
        link
        fedilink
        English
        arrow-up
        2
        ·
        1 year ago

        We need something like the fair use doctrine coupled with identify rights.

        If you want to use X’s voice and likeness in something, you have to purchase that privilege from X or X’s estate, and they can tell you to pay them massive fees or to fuck off.

        Fair use would be exclusively for comedy, but still face regulation. There’s plenty of hilarious TikToks that use AI to make characters say stupid shit, but we can find a way to protect voice actors and creators without stifling creativity. Fair use would still require the person’s permission, you just wouldn’t need to pay to use it for such a minor thing – a meme of Mickey Mouse saying fuck for example.

        At the end of the day though, people need to hold the exclusive and ultimate right to how their likeness and voice are used, and they need to be able to shut down anything they deem unacceptable. Too many people are concerned with what is capable than with acting like an asshole. It’s just common kindness to ask someone if you can use their voice for something, and respecting their wishes if they don’t want it.

        I don’t know if this is a hot take or not, but I’ll stand by it either way – using AI to emulate someone without their permission is a fundamental violation of their rights and privacy. If OpenAI or whoever wants to claim that makes their product unusable, tough fucking luck. Every technology has faced regulations to maintain our rights, and if a company can’t survive without unbridled regulations, it deserves to die.

    • Jaded@lemmy.dbzer0.com
      link
      fedilink
      English
      arrow-up
      14
      arrow-down
      14
      ·
      1 year ago

      What about the third option, everyone gets to have the power?

      I’ve seen what Marvin Gaye and Conan Doyle’s relatives have done with the power. Dump it in the creative commons. Nobody should own the tonalities of a voice anyways, there quickly wouldn’t be any left.

      • Puzzle_Sluts_4Ever@lemmy.world
        link
        fedilink
        English
        arrow-up
        21
        arrow-down
        6
        ·
        1 year ago

        Considering the internet is already a hellscape of deepfake porn, let’s not take the libertarian approach to this, 'kay?

        Also, there are two major issues at hand that you are conflating.

        People aren’t doing AI recreations of Robin Williams because they love the way he said “zucchini”. They are doing it because of the novelty of hearing Robin perform their material or making him say “Happy Birthday Fred” or “Jewish Space Lizards Control Kansas” or whatever. Much like with deepfake porn, the appeal is using someone against their will for your own pleasure.

        The other aspect, and what the SAG and WGA strikes have been about (and which Robin famously preempted over twenty years ago), is training data. It is the idea of using past footage and performances to make a super actor (similar to what Square tried with FF The Spirits Within). So you might have Tom Cruise’s gait coupled with Ryan Reynolds’s chin and Hugh Jackman’s nipples and so forth. And, that is still a huge mess.

          • Puzzle_Sluts_4Ever@lemmy.world
            link
            fedilink
            English
            arrow-up
            10
            arrow-down
            5
            ·
            1 year ago

            If your bad faith requirement is complete eradication, sure.

            If the goal is to vastly diminish the amount of content out there by preventing monetization and providing a legal means to pull said content? As well as to vilify the concept? Then yeah, it works.

          • wanderingmagus@lemm.ee
            link
            fedilink
            English
            arrow-up
            4
            arrow-down
            1
            ·
            1 year ago

            Prohibition of CSAM seems to be universally accepted as a thing we should keep doing. What say you to that?

      • brsrklf@jlai.lu
        link
        fedilink
        English
        arrow-up
        3
        ·
        1 year ago

        In the context of close relatives being very disturbed by what is made with the person’s image, I really don’t think legally allowing absolutely everyone to do as they please with it will help.