4 ms·

For me there were two key parts of this post: 1. On prediction: If you had a maximally detailed model of someone's worldview, you would be able to predict thei

For me there were two key parts of this post: 1. On prediction: If you had a maximally detailed model of someone's worldview, you would be able to predict their views accurately at a given time. I'm sympathetic to this in principle, because people are often very predictable and stick within their rubric. I would add, though, that worldviews change and evolve in weird ways, and sometimes people also have views that ... don't seem entirely predetermined by their worldviews or their own history? I think there can be something situational and emergent about our views, and this isn't completely predetermined by a worldview. (Let's leave aside that we don't have enough data to accurately construct the originally suggested model — we aren't omniscient, but I for one am OK with that.) 2. On hidden premises: Sometimes you try to have a conversation with someone and find out that they have some basic differences in their theory of knowledge such that you can't get through to them at all. Like "No matter what you say is true, I will hypothesize that evil hidden forces are misleading us." (It's kind of the postmodern version of Descartes' evil demon hypothesis, if you like.) I don't find this claim too surprising in the general case, but I do think, in the specific political context of the USA at the moment, it's worth remembering that it is a nation divided not only by political but also by epistemological differences. It's a long post but these were the parts I found the most interesting.

mandmandam 2y ago
1. This was a hypothetical situation where perfect prediction were possible, explicitly stated as an assumption. Not as fact, but as the basis of a thought experiment. 2. I'm not sure what part of the article you're referring to here, because as far as I understand it, the author is speaking about political and epistemological and every other possible difference, framed as axes in a worldview space.
- smeej 2y ago
  Not the person you're replying to, but I rolled my eyes and had trouble taking the rest of the article seriously when such an implausible thought experiment came so early in the article without being criticized for the deep flaws in this assumption. I don't think anyone is internally consistent all the time (yes, including me). Whatever principle or set of principles someone thinks they have, there's always some conceivable situation where we could line them all up so there should be an obvious "right" conclusion, but the person will still instinctively react against it and feel it's wrong. This is why we won't ever be able to, for example, train up AI surrogates to go have all our arguments for us, on a societal level, faster and in more depth than we ever could since we have to worry about things like continuing to eat, and let them make decisions about agreements that should govern our actions. The agents would run into our self-contradictions all the time, and have to point them out to us to try to get clarity, and then we'd get so frustrated with their pointing this out all the time that we'd turn them off and abandon the whole endeavor.
  - mandmandam 2y ago
    > I rolled my eyes and had trouble taking the rest of the article seriously when such an implausible thought experiment came so early in the article without being criticized for the deep flaws in this assumption The thought experiment wasn't supposed to be 'plausible'. It's purpose was to introduce the concept of worldviews residing in a high dimensional space, which could perhaps be approximated. It was also actually a quote from an earlier article, with more context. > I don't think anyone is internally consistent all the time I agree. Only a moron or a God could accomplish such a feat. However, this doesn't invalidate the thought experiment - all you need to do is add a time axis. "On this particular morning, the Judge had dreamed that she sat in a beautiful parlor while doves fed grain to her kittens. Her general amiability is notably up, and she will deliver far less lenient verdicts all morning because of her contrarian tendencies. Then, a bone in her smoked haddock sandwich triggers repressed childhood memories, leaving her 31.64% more disaffected and apathetic." - The thought experiment could encompass full knowledge of all of this, from moment to moment. > This is why we won't ever be able to, for example, train up AI surrogates to go have all our arguments for us, on a societal level ... and let them make decisions about agreements that should govern our actions Is that why? Or is it because by the time AI is that advanced the likelihood of our being able to control it is virtually nil, thanks to a breakneck developmental arms race... Besides which, being frustrated with clarity is something of a contradiction in itself, lol. I imagine a hyper-intelligent being could teach us to be better people in a way we'd actually like, if they were into that sort of thing. The Culture books explore this quite well.
  - punpunia 2y ago
    Since our actions, thoughts are results of complex physical systems within our body, trying to abstract away to some worldview space will always be just that, an abstraction. The only space in which we could make accurate predictions as to a persons viewpoints in different situations is within the mind of Laplace's Demon.
  - DeathArrow 2y ago
    >This is why we won't ever be able to, for example, train up AI surrogates to go have all our arguments for us Some AI models like those from OpenAI or Anthropic are biased to the left. Anyone can test that. So actually people thought asking machines build arguments for them and trained them accordingly.
keiferski 2y ago
I don’t see how a maximally detailed model of someone’s worldview isn’t just a list of their views. Or at least it would be inclusive of their views. And so by definition you’d be able to predict them, as you already have them.

mandmandam 2y ago

1. This was a hypothetical situation where perfect prediction were possible, explicitly stated as an assumption. Not as fact, but as the basis of a thought experiment. 2. I'm not sure what part of the article you're referring to here, because as far as I understand it, the author is speaking about political and epistemological and every other possible difference, framed as axes in a worldview space.

smeej 2y ago
Not the person you're replying to, but I rolled my eyes and had trouble taking the rest of the article seriously when such an implausible thought experiment came so early in the article without being criticized for the deep flaws in this assumption. I don't think anyone is internally consistent all the time (yes, including me). Whatever principle or set of principles someone thinks they have, there's always some conceivable situation where we could line them all up so there should be an obvious "right" conclusion, but the person will still instinctively react against it and feel it's wrong. This is why we won't ever be able to, for example, train up AI surrogates to go have all our arguments for us, on a societal level, faster and in more depth than we ever could since we have to worry about things like continuing to eat, and let them make decisions about agreements that should govern our actions. The agents would run into our self-contradictions all the time, and have to point them out to us to try to get clarity, and then we'd get so frustrated with their pointing this out all the time that we'd turn them off and abandon the whole endeavor.
- mandmandam 2y ago
  > I rolled my eyes and had trouble taking the rest of the article seriously when such an implausible thought experiment came so early in the article without being criticized for the deep flaws in this assumption The thought experiment wasn't supposed to be 'plausible'. It's purpose was to introduce the concept of worldviews residing in a high dimensional space, which could perhaps be approximated. It was also actually a quote from an earlier article, with more context. > I don't think anyone is internally consistent all the time I agree. Only a moron or a God could accomplish such a feat. However, this doesn't invalidate the thought experiment - all you need to do is add a time axis. "On this particular morning, the Judge had dreamed that she sat in a beautiful parlor while doves fed grain to her kittens. Her general amiability is notably up, and she will deliver far less lenient verdicts all morning because of her contrarian tendencies. Then, a bone in her smoked haddock sandwich triggers repressed childhood memories, leaving her 31.64% more disaffected and apathetic." - The thought experiment could encompass full knowledge of all of this, from moment to moment. > This is why we won't ever be able to, for example, train up AI surrogates to go have all our arguments for us, on a societal level ... and let them make decisions about agreements that should govern our actions Is that why? Or is it because by the time AI is that advanced the likelihood of our being able to control it is virtually nil, thanks to a breakneck developmental arms race... Besides which, being frustrated with clarity is something of a contradiction in itself, lol. I imagine a hyper-intelligent being could teach us to be better people in a way we'd actually like, if they were into that sort of thing. The Culture books explore this quite well.
- punpunia 2y ago
  Since our actions, thoughts are results of complex physical systems within our body, trying to abstract away to some worldview space will always be just that, an abstraction. The only space in which we could make accurate predictions as to a persons viewpoints in different situations is within the mind of Laplace's Demon.
- DeathArrow 2y ago
  >This is why we won't ever be able to, for example, train up AI surrogates to go have all our arguments for us Some AI models like those from OpenAI or Anthropic are biased to the left. Anyone can test that. So actually people thought asking machines build arguments for them and trained them accordingly.

keiferski 2y ago

I don’t see how a maximally detailed model of someone’s worldview isn’t just a list of their views. Or at least it would be inclusive of their views. And so by definition you’d be able to predict them, as you already have them.