ControlsFreak
No bio...
User ID: 1422
ControlsFreak
2d ago
There's a pretty huge difference between a tape of someone running their mouth and a tape of someone raping a minor. Again, if your model of the world doesn't account for this sort of massive difference, then you might want to reconsider your model. Different models may have different predictions for a tape of someone running their mouth, and one might evaluate said models on what actually happened, but there is obviously no constraint on the set of models forcing them to produce the same output on such extremely different cases.
people would say it's AI
This is plausible today, which is why I mentioned it.
they'd think it was out of context roleplay ... they'd say she lied about her age ... they'd think it was invasion of privacy or propaganda and refuse to watch ... they'd think Trump has let himself down again, but on a national level he's still a force for good etc
None of these are plausible for the example given of a tape of him "violently raping a 14-year-old girl".
Testing one's model as parameters go to infinity is, indeed, a good sanity check. I do this in my daily work. If your model has truly absurd results as the parameters go to infinity, it's more likely that there's a problem with your model than that the world will actually match the model outputs.
ControlsFreak
2d ago
This is one of those moments where you should probably take honest stock in your model of the world, because it's really far out there. I could imagine some defenses these days along the lines of the video not being real; AI gen has gotten good or whatever. But there is not even a single cultural/theoretical/whathaveyou hook that is remotely likely to take hold as a defense in society if it is widely believed that such a video is real. It's not like Clinton, where the left was already trying to lean hard on "consent of adults is the only thing that matters" in order to help the gays.
ControlsFreak
3d ago
I did say that I was sure I would link to SMBC doing the philosophy of mathematics joke many times in the future here.
ControlsFreak
3d ago
If you are definitionally not allowed to observe an empirical difference then the answer to the question is mu, as both answers yield exactly identical predictions about the future and so are the same answer.
This is a fairly common failure in reasoning from STEM people who haven't STEMed enough. You may just be unfamiliar with the concept of observability. That's not even getting into the actual philosophy problem.
The maths fail part of the STEM fail has already been covered decently enough below.
It's all just fancy window dressing over consequentialist reasoning.
This, on the other hand, isn't a STEM fail; it is definitely outside of that. But it does give me yet another chance to share one of my favorite papers on the topic.
Sure, some people will want to think that it's fake in some way. I mean, I guess something like that could be fake? If you asked me a prior probability for a video coming out of any prominent politician committing a violent rape of a 14-year-old, especially in the AI age, I'd have a pretty non-zero chance of it being fake. And these days, normies have had their probability estimates for foreign government disinformation along lines like these jacked up, too.
...but that's basically the only thing that could plausibly have any play for the example given. People might think it's fake, but if there is enough other evidence to support that it's not a total fabrication, nothing else would save him.
More options
Context Copy link