I think your argument applies well to misinformation/deepfakes, but not to sex/relationships/pornography, on the basis that they are targeted towards two different orders of needs. When seeking information, we care a lot about veracity and usefulness, and the value gained by small tweaks to an existing story is counteracted by the value lost by that story moving from ‘entirely true’ to ‘not entirely true’. As an example, if there was a terrorist attack and twelve people died, then an AI edit of the story and related media that claims thirteen people died would not be straightforwardly better than the original in terms of ability to propagate or be capitalized upon. AI provides the ability to A/B test at scale and cheaply personalize, which is of relatively limited usefulness in that space.
In contrast, if I’m in the adult video business, I’m targeting the lizard brain. If an AI generated video has some immediate visual features that hit better than real ones, then my video will get more clicks, as users look for the content that elicits the strongest reaction. There isn’t a fundamental limit to this, either—plenty of people already make money by producing completely animated AVs, for example, which definitely don’t pass for real footage. Moreover, there’s currently a tradeoff between actor quality and novelty - ‘weirder’ fetishes that are specific to small groups are less able to attract the most or the best performers, which serves to mitigate the “novelty spirals” that we constantly hear about, hitting a fundamental upper bound when a fetish is so specific to a given individual that nobody at all has produced a video that services it.
At a bare minimum, I think that AI-generated videos will outcompete real pornographic content and take the limits off of divergence from “normal” sexual tastes among heavier consumers. The gradient towards weirder content can become infinitely small, mitigating the shock of a piece of content that goes “too far” for someone’s current preferences, and every performer can be a flawless match to the viewer’s exact tastes.
I think your argument applies well to misinformation/deepfakes, but not to sex/relationships/pornography, on the basis that they are targeted towards two different orders of needs. When seeking information, we care a lot about veracity and usefulness, and the value gained by small tweaks to an existing story is counteracted by the value lost by that story moving from ‘entirely true’ to ‘not entirely true’. As an example, if there was a terrorist attack and twelve people died, then an AI edit of the story and related media that claims thirteen people died would not be straightforwardly better than the original in terms of ability to propagate or be capitalized upon. AI provides the ability to A/B test at scale and cheaply personalize, which is of relatively limited usefulness in that space.
In contrast, if I’m in the adult video business, I’m targeting the lizard brain. If an AI generated video has some immediate visual features that hit better than real ones, then my video will get more clicks, as users look for the content that elicits the strongest reaction. There isn’t a fundamental limit to this, either—plenty of people already make money by producing completely animated AVs, for example, which definitely don’t pass for real footage. Moreover, there’s currently a tradeoff between actor quality and novelty - ‘weirder’ fetishes that are specific to small groups are less able to attract the most or the best performers, which serves to mitigate the “novelty spirals” that we constantly hear about, hitting a fundamental upper bound when a fetish is so specific to a given individual that nobody at all has produced a video that services it.
At a bare minimum, I think that AI-generated videos will outcompete real pornographic content and take the limits off of divergence from “normal” sexual tastes among heavier consumers. The gradient towards weirder content can become infinitely small, mitigating the shock of a piece of content that goes “too far” for someone’s current preferences, and every performer can be a flawless match to the viewer’s exact tastes.