If you want a post explaining the same concepts to a different audience, then go write a post explaining the same concepts to a different audience. I am well aware of the tradeoffs I chose here. I wrote the post for a specific purpose, and the tradeoffs chosen were correct for that purpose.
On the one hand, yeah, that's the dream.
On the other hand, focusing on people and groups and working together seems to be the #1 way that people lose track of wizard power in practice, and end up not having any. It's just so much easier to say to oneself "well, this seems hard, but maybe I can get other people in a group to do it", and to do that every time something nontrivial comes up, and for most people in the group to do it every time something comes up, until most of what the group actually does is play hot potato while constructing a narrative about how valuable it is for all these people to be working together.
I don't know the full sequence of things such a person needs to learn, but probably the symbol/referent confusion thing is one of the main pieces. The linked piece talks about it in the context of "corrigibility", but it's very much the same for "consciousness".
Yeah, Stephen's comment is indeed a mild update back in the happy direction.
I'm still digesting, but a tentative part of my model here is that it's similar to what typically happens to people in charge of large organizations. I.e. they accidentally create selection pressures which surround them with flunkies who display what the person in charge wants to see, and thereby lose the ability to see reality. And that's not something which just happens to crazies. For instance, this is my central model of why Putin invaded Ukraine.
None of that about AI relationships sounds particularly bad. Certainly that's not the sort of problem I'm mainly worried about here.
That... um... I had a shortform just last week saying that it feels like most people making heavy use of LLMs are going backwards rather than forwards. But if you're getting 10-20 of that per day, and that's just on LessWrong... then the sort of people who seemed to me to be going backward are in fact probably the upper end of the distribution.
Guys, something is really really wrong with how these things interact with human minds. Like, I'm starting to think this is maybe less of a "we need to figure out the right ways to use the things" sort of situation and more of a "seal it in a box and do not touch it until somebody wearing a hazmat suit has figured out what's going on" sort of situation. I'm not saying I've fully updated to that view yet, but it's now explicitly in my hypothesis space.
Set , then note that the stochastic error ) because induces perfect conditional independence and symmetry of and .
I don't think Y induces perfect conditional independence? Conditional on Y, we have:
... which means that learning the value of A tells me something about the value of B, conditional on Y (specifically, B is more likely to have the same value A had).
Am I missing something here?
(Also, for purposes of me tracking how useful LLMs are for research: assuming I'm not missing something and this was a mistake, was the mistake originally made by you or an LLM?)
Can you give an example of what a "most fun" conversation looked like? What's the context, how did it start, how did the bulk of it go, how did you feel internally throughout, and what can you articulate about what made it so great?
High vs low voltage has very different semantics at different places on a computer chip. In one spot, a high voltage might indicate a number is odd rather than even. In another spot, a high voltage might indicate a number is positive rather than negative. In another spot, it might indicate a jump instruction rather than an add.
Likewise, the same chemical species have very different semantics at different places in the human body. For example, high serotonin concentration along the digestive tract is a signal to digest, whereas high serotonin concentration in various parts of the brain signals... uh... other stuff. Similarly, acetylcholine is used as a neurotransmitter both at neuromuscular junctions and in the brain, and these have different semantics. More generally, IIUC neurotransmitters like dopamine, norepinephrine, or serotonin are released by neurons originating at multiple anatomically distinct little sub-organs in the brain. Each sub-organ projects to different places, and the same neurotransmitter probably has different semantics when different sub-organs project to different targets.
Yet most pharmaceutical interventions target one type of molecule, or one receptor, or what have you, approximately everywhere. Such an intervention is analogous to e.g. attempting to make every float in a computer's memory positive by flipping the first bit in every block, but then as a side-effect also changing a bunch of jump instructions to add instructions because there was no way to localize the effect to float-containing memory locations.
Thus the question: how can biochemical interventions be localized, especially in general-purpose ways? I'll throw out some ideas off the top of my head, but I'm interested to hear other peoples' thoughts as well.
The blood-brain barrier springs to mind as one example. If a chemical has different semantics in the brain and outside, and one wishes to target outside the brain, then just use a drug which can't cross the barrier.
One could put an implant in the right spot to release a drug, and then choose a drug which either isn't transported quickly or breaks down before it can get very far (or both).
Notably, making some random molecule diffuse less quickly seems relatively tractable: one can just attach a bigger molecule to it. And there's an absolutely enormous space of possibilities for what that bigger molecule could be, so it's especially likely to be tractable.
Cells already need the ability to tell "where they are" in order for us to have anatomically distinct regions at all. So in principle, it should be possible to genetically modify cells to do something different, but gate the change on the cell being in a particular distinct anatomical region, so cells everywhere else do the same thing as before.
For adult genetic modifications, one would probably want to combine this method with something similar to the implant + slow transport/fast release method above. Adult genetic modifications usually don't hit every cell or even a majority of them, so an ideal use would be modifying some small percentage of cells to release a molecule which influences all the others. Slow diffusion/fast breakdown could then localize that molecule.
I'm curious about other methods to localize biochemical interventions in the body, both speculative and already-existing.
Agreed, I don't think it's actually that rare. The rare part is the common knowledge and normalization, which makes it so much easier to raise as a hypothesis in the heat of the moment.