I think Eliezer once wrote something about things becoming clearer when you think about how you would program a computer to do it, as opposed to e.g. just throwing some applause lights to a human. So, how specifically would you implement this kind of belief in a computer?
Also, should we go meta and say: "'Rationality gives us a better understanding of the world, except when it does not' is a good ideology, except when it is worse" et cetera?
What exactly would that actually mean? (Other than verbally shielding yourself from criticism by endless "but I said 'except when not'".) Suppose a person A believes "there is a 80% probability it will rain tomorrow", but a person B believes "there is a 80% probability it will rain tomorrow, except if it is some different probability". I have an idea about how A would bet about tomorrow's weather, but how would B?
I think Eliezer once wrote something about things becoming clearer when you think about how you would program a computer to do it, as opposed to e.g. just throwing some applause lights to a human. So, how specifically would you implement this kind of belief in a computer?
First solve natural language...
No one has used eliezer's technique much and there may be a reason for that.
Operating outside of ideology is extremely hard, if not impossible. Even groups that see themselves as non-ideological, still seem to end up operating within an ideology of some sort.
Take for example Less Wrong. It seems to operate within a few assumptions:
...
These assumptions are also subject to some criticisms. Here's one criticism for each of the previous points:
I could continue discussing assumptions and possible criticisms, but that would be a distraction from the core point, which is that there are advantages to having a concrete ideology that is aware of it's own limitations, as opposed to an implicit ideology that is beyond all criticism.
Self-conscious ideologies also have other advantages: