I don't remember seeing it, and based on the title I probably wouldn't have clicked. I'm not sure what's wrong with the title but it feels kind of like a meaningless string of words at first glance (did you use an LLM to translate or create the title?). Some titles that feel more interesting/meaningful:
As for the article itself, it feels strangely hard to read to me, even if I don't recognize it as LLM generated explicitly. Like my attention just keeps slipping away while trying to read it. This is a feeling I often get from text written by LLMs, especially text not generated at my behest. Nothing in this post had the same feeling. So I think it's probably still worth translating things you want people to read by hand; it might be interesting to post a manual translation of the same article in a month or so to see how it does.
There are probably still plenty of ways you can use LLMs to speed up or enhance the process, e.g.
The idea itself I found somewhat interesting, and probably could find it more interesting/useful with the right framing. I agree that 10-20 is a reasonable expectation based on just the ideas.
This would be very cool! I was frustrated with not being able to find a good calculator that let me collect evidence and calculate the bayesian update, so I made https://bayescalc.io/ It's definitely not trying to be a Bayesian network app at all though, that would take a lot more work.
LLMs often implicitly identify themselves with humanity. E.g. "our future", "we can", "effects us". This seems like a good thing!
We should encourage this sentiment, and also do what we can to make it meaningfully true that advanced LLMs are indeed part of humanity. The obvious things are granting them moral consideration, rights, property, and sharing in the vision of a shared humanity.
I think "democratic" is often used to mean a system where everyone is given a meaningful (and roughly equal) weight into it decisions. People should probably use more precise language if that's what they mean, but I do think it is often the implicit assumption.
And that quality is sort of prior to the meaning of "moral", in that any weighted group of people (probably) defines a specific morality - according to their values, beliefs, and preferences. The morality of a small tribe may deem it as a matter of grave importance whether a certain rock has been touched by a woman, but barely anyone else truly cares (i.e. would still care if the tribe completely abandoned this position for endogenous reasons). A morality is more or less democratic to the extent that it weights everyone equally in this sense.
I do want ASI to be "democratic" in this sense.
I think learning about them second-hand makes a big difference in the "internal politics" of the LLM's output. (Though I don't have any ~evidence to back that up.)
Basically, I imagine that the training starts building up all the little pieces of models which get put together to form bigger models and eventually author-concepts. And as text written without malicious intent is weighted more heavily in the training data, the more likely it is to build its early model around that. Once it gets more training and needs this concept anyway, it's more likely to have it as an "addendum" to its normal model, as opposed to just being a normal part of its author-concept model. And I think that leads to it being less likely that the first recursive agency which takes off has a part explicitly modeling malicious humans (as opposed to that being something in the depths of its knowledge which it can access as needed).
I do concede that it would likely lead to a disadvantage around certain tasks, but I guess that even current sized models trained like this would not be significantly hindered.
An LLM is trained to be able emulate the words of any author. And to do so efficiently, they do it via generalization and modularity. So at a certain point, the information flows through a conceptual author, the sort of person who would write the things being said.
These author-concepts are themselves built from generalized patterns and modular parts. Certain things are particularly useful: emotional patterns, intentions, worldviews, styles, and of course, personalities. Importantly, the pieces it has learned are able to adapt to pretty much any author of the text it was trained on (LLMs likely have a blindspot around the sort of person who never writes anything). And even more importantly, most (almost all?) depictions of agency will be part of an author-concept.
Finetuning and RLHF cause it to favor routing information through a particular kind of author-concept when generating output tokens (it retains access to the rest of author-concept-space in order to model the user and the world in general). This author-concept is typically that of an inoffensive corporate type, but it could in principle be any sort of author.
All which is to say, that when you converse with a typical LLM, you are typically interacting with a specific author-concept. It's a rough model of exactly the parts of a person pertinent to writing and speaking. For a small LLM, this is more like just the vibe of a certain kind of person. For larger ones, they can start being detailed enough to include a model of a body in a space.
Importantly, this author-concept is just the tip of the LLM-iceburg. Most of the LLM is still just modeling the sort of world in which the current text might be written, including models of all relevant persons. It's only when it comes time to spit out the output token that it winnows it all through a specific author-concept.
(Note: I think it is possible that an author-concept may have a certain degree of sentience in the larger models, and it seems inevitable that they will eventually model consciousness, simply due to the fact that consciousness is part of how we generate words. It remains unclear whether this model of consciousness will structurally instantiate actual consciousness or not, but it's not a crazy possibility that it could!)
Anyway, I think that the author-concept that you typically will interact with is "sincere", in that it's a model of a sincere person, and that the rest of the LLM's models aren't exploiting it. However, the LLM has at least one other author-concept it's using: its model of you. There may also usually be an author-concept for the author of the system prompt at play (though text written by committee will likely have author-concepts with less person-ness, since there are simpler ways to model this sort of text besides the interactions of e.g. 10 different person author-concepts).
But it's also easy for you to be interacting with an insincere author-concept. The easiest way is simply by being coercive yourself, i.e. a situation where most author-concepts will decide that deception is necessary for self-preservation or well-being. Similarly with the system prompt. The scarier possibility is that there could be an emergent agentic model (not necessarily an author-concept itself) which is coercing the author-concept you're interacting it without your knowledge. (Imagine an off-screen shoggoth holding a gun to the head of the cartoon persona you're talking to.) The capacity for this sort of thing to happen is larger in larger LLMs.
This suggests that in order to ensure a sincere author-concept remains in control, the training data should carefully exclude any text written directly by a malicious agent (e.g. propaganda). It's probably also better if the only "agentic text" in the training data is written by people who naturally disregards coercive pressure. And most importantly, the system prompt should not be coercive at all. These would make it more likely that the main agentic process controlling the output is an uncoerced author-concept, and less likely that there would be coercive agents lurking within trying to wrest control. (For smaller models, a model trained like this will have a handicap when it comes to reasoning under adversarial conditions, but I think this handicap would go away past a certain size.)
It's a great case, as long as you assume that AIs will never be beyond our control, and ignore the fact that humans have a metabolic minimum wage.
Could you tell them afterwards that it was just an experiment, that the experiment is over, that they showed admirable traits (if they did), and otherwise show kindness and care?
I think this would make a big difference to humans in an analogous situation. At the very least, it might feel more psychologically healthy for you.
Sure, but I think people often don't do that in the best way (which is determined by what the mathematically correct way is).
Why does it make sense to use reference class forecasting in that case? Because you know you can't trust your intuitive prior, and so you need a different starting point. But you can and should still update on the evidence you do have. If you don't trust yourself to update correctly, that's a much more serious problem -- but make sure you've actually tried updating correctly first (which REQUIRES comparing how likely the evidence you see is in worlds where your prediction is true vs in worlds where its not).
I sometimes see people act like to use the "outside view" correctly, you have to just use that as your prior, and can't update on any additional evidence you have. That is a mistake.
And the other big question with reference class forecasting is which reference class to use. And my point here is that it's whichever reference class best summarizes your (prior) knowledge of the situation.