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.)
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.
I don't disagree that totalitarian AI would be real bad. It's quite plausible to me that the "global pause" crowd are underweighting how bad it would be.
I think an important crux here is on how bad a totalitarian AI would be compared to a completely unaligned AI. If you expect a totalitarian AI to be enough of an s-risk that it is something like 10 times worse than an AI that just wipes everything out, then racing starts making a lot more sense.
I think mostly we're on the same page then? Parents should have strong rights here, and the state should not.
I think that there's enough variance within individuals that my rule does not practically restrict genomic liberty much, while making it much more palatable to the average person. But maybe that's wrong, or it still isn't worth the cost.
Your rule might for example practically prevent a deaf couple from intentionally having a child who is deaf but otherwise normal. E.g. imagine if the couple's deafness alleles also carry separate health risks, but there are other deafness alleles that the couple does not have but that lead to deafness without other health risks.
That's a good point, I wouldn't want to prevent that. I'm not sure how likely this is to practically come up though.
Restrictions on genomic liberty should be considered very costly: they break down walls against eugenics-type forces (i.e. forces on people's reproduction coming from state/collective power, and/or aimed at population targets).
Strong agree.
However, the difference is especially salient because the person deciding isn't the person that has to live with said genes. The two people may have different moral philosophies and/or different risk preferences.
A good rule might be that the parents can only select alleles that one or the other of them have, and also have the right to do so as they choose, under the principle that they have lived with it. (Maybe with an exception for the unambiguously bad alleles, though even in that case it's unlikely that all four of the parent's alleles are the deleterious one or that the parents would want to select it.) Having the right to select helps protect from society/govt imposing certain traits as more or less valuable, and keeping within the parent's alleles maintains inheritance, which I think are two of the most important things people opposed to this sort of thing want to protect.
Thank you for doing this research, and for honoring the commitments.
I'm very happy to hear that Anthropic has a Model Welfare program. Do any of the other major labs have comparable positions?
To be clear, I expect that compensating AIs for revealing misalignment and for working for us without causing problems only works in a subset of worlds and requires somewhat specific assumptions about the misalignment. However, I nonetheless think that a well-implemented and credible approach for paying AIs in this way is quite valuable. I hope that AI companies and other actors experiment with making credible deals with AIs, attempt to set a precedent for following through, and consider setting up institutional or legal mechanisms for making and following through on deals.
I very much hope someone makes this institution exist! It could also serve as an independent model welfare organization, potentially. Any specific experiments you would like to see?
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.