The primary virtue is control

How do we perceive The Good? It seems like we get an intuitive sense of Goodness when looking at Good things. How do we know that this is a detector of the real Goodness, like our eyes are detectors of the real world, and not just "hardcoded", only checking our own internal sense of Goodness?

At which point in the causal history of your brain did The Good influence your brain to be able to detect it, as opposed to mistakenly detecting something else as The Good?

If you built a creature that had its own internal Good detector tuned to something random, how would it be able to tell that wasn't actually Good and orient itself towards the real Good? I dont think it can. Are we in the same situation as this creature?

Does The Good affect anything physical? If so, how? If not, how did your physical brain learn what it is?

Sorry for the long ask. Questions are not intended as patronising, ive never really heard the moral realist side.

arguing from evolution towards the reality of the external world is totally circular! we believe evolution is true because of our sense data!

we *dont* actually know that our eyes are detectors of the real world! we make a leap of faith, of assumption! try to stop doing science for a second, and do philosophy!

but also thinking of it as "detectors" isn't right at all, maybe my metaphor wasn't entirely clear....

i am trying to express a radically different way of thinking about your mind and what it does. that it doesn't just... like its not actually some sort of logical deduction-machine. its a thing that assess probability/possibility/truth-value in a sort of deeply heuristic way. that forms convictions, that "judges plausibility". and that the existence of good things and bad things is felt with a certainty and intensity as certain and intense as the way your sense data give you the information.

i think this epistemology of mine kind of "allows for" delusions in a way that is not typical. but i think EVERY epistemology allows for delusions in the same way, and just ignores this fact. like, yes, you can choose to ignore that we (you!) are the type of being that experiences these sort of weird intense convictions "out of nowhere". but you shouldnt! i think the system that gives you those delusions is the same system that gives you anything else! they cant be removed from the system!

and then... all your latter questions come from this conception where youre imagining im saying the good is like...a physical object or force? or something? thats not what i mean. like, imagine if i said "there is such a thing as the number 2, it defilynitely exists" and you were like "how do we perceive 2", "what if you had your 2 detector tuned wrong", "does 2 affect anything physical"? its nonsense! i mean maybe 2 isnt the best example because like, you could say we just "define" 2. but there were 2 of things before we defined 2! we didnt like, invent 2-ness, there were actually quantities of things, specifically 2 of them!

What i mean by "reality" or "detecting vs hardcoding" is if, basically, different people (broadly construed) see the same thing. Imagine a system of a bunch of different people, in an environment. Some of the people have "environment detectors", such that their mental states correlate with their environments. Other times, the people have their senses "hardcoded", what they see doesn't depend on the environment at all. The people cannot, themselves, tell what type they are. They have Cartesian doubt. However, if you have access to all the mental states but not the environment, you can tell who is detecting and who is blind because the mental states of all the detectors are correlated with one another, because they are each correlated with the environment.

This is what I mean by "detecting reality". A number of different types of minds (people and animals are basically all we have) all seem to detect the same thing, external to them. The reference to evolution comes up because reality detecting beings have higher fitness than beings with hardcoded senses.

Now, say these people that we watching have a deeply felt intuitive sense of The Good. How could we tell if these senses were hardcoded, or detecting? Compare what different people say about The Good, and check if they're correlated. When I say other people, comparing humans doesnt count, because our minds are already correlated. If we had insight into the mind of an animal, and it considered a lot of the same things as us Good, without a reason for why that correlation exists, I would consider that evidence of moral realism! In a sense, we cant really answer this question until we find out what total aliens think of The Good. If they came to the same conclusion as us, we probably both have Good detectors. Otherwise, we probably have hardcoded senses of The Good.

As for 2, I expect that many different types of mind would have conceptions of 2, because 2 and its properties are present in the real world. When you have 2 rocks, and you take one away, and then another one, you have no rocks left. This could be said to be 2 "influencing" the real world in a way that we can detect, or probably more accurately "the world exhibiting 2-structure". If you had 2 rocks, and took away one, leaving you with 3 rocks, and took away another, leaving you with 10 rocks, and took away another, leaving you with no rocks, and we had observed this fact all through history, and all things which currently follow 2-structure now follow this structure, i dont actually think that people living in that world would know what 2 is. They'd have no way to, because they have no 2-detectors. Our general reality detectors work as 2-detectors because reality has 2-structure. See also https://www.lesswrong.com/posts/6FmqiAgS8h4EJm86s/how-to-convince-me-that-2-2-3.

this paper is amazing

they modify an LLM to attend over an array of token streams instead of a single one. with a ton of interesting consequences:

user input and the model's chain of thought can happen in parallel

user input and the model's response can happen in parallel, allowing them to interrupt each other

you can have multiple parallel chains of thought running at the same time!

you can separate out the system prompt into its own token stream, and theoretically train the model to detect and ignore prompt injection this way

so that's all very paradigm shifty and cool but also these chains of thought are sooooooooo cute. look at these!