Didn’t read the article. Why? Don’t care enough what the author thinks.
I’ve considered this before. IMO, it’s kind of a fruit from the poison tree situation. If all porn in the model is consensual and legal, then it actually has the capacity to significantly undercut an industry that has a lot of exploitation in it (particularly the illegal, inconsensual part), because now you can just create what you want to see when you want to see it, and that seems like it’s a lot easier and cheaper than paying a criminal or group of criminals to hurt someone. The reality, of course, is that the model almost certainly has ingested illegal or inconsensual content. Ethically speaking, that could mean that anything it produces will carry that stain with it. I think there’s a potential here to reduce net suffering, but it’s like any tool; it all depends on how people decide to use it, and I’m not sure that the bad actors won’t just completely ruin this for everyone.
In your hypothesis, it is assumed that the model is trained in some small part on unethical content, but where your logic deviates is then labeling the product of that training unethical when said exploitation has not been perpetuated or renewed. By the same judgment, ethical actions are impossible after the historical first (and “original sin” is systemic crowd control, not cultural metric).
Didn’t read the article. Why? Don’t care enough what the author thinks.
I’ve considered this before. IMO, it’s kind of a fruit from the poison tree situation. If all porn in the model is consensual and legal, then it actually has the capacity to significantly undercut an industry that has a lot of exploitation in it (particularly the illegal, inconsensual part), because now you can just create what you want to see when you want to see it, and that seems like it’s a lot easier and cheaper than paying a criminal or group of criminals to hurt someone. The reality, of course, is that the model almost certainly has ingested illegal or inconsensual content. Ethically speaking, that could mean that anything it produces will carry that stain with it. I think there’s a potential here to reduce net suffering, but it’s like any tool; it all depends on how people decide to use it, and I’m not sure that the bad actors won’t just completely ruin this for everyone.
In your hypothesis, it is assumed that the model is trained in some small part on unethical content, but where your logic deviates is then labeling the product of that training unethical when said exploitation has not been perpetuated or renewed. By the same judgment, ethical actions are impossible after the historical first (and “original sin” is systemic crowd control, not cultural metric).