I was a physics undergrad, but I did an AI PhD on artificial learners adopting normative conventions from human teachers. This work included an attempt to formalise the learning situation, where an artificial learner, tries to figure out, what one human teacher, would like that learner to do. Later, I also worked on interpreting human feedback. Here is a review article, that situates these two strands of research in a larger context. This review article also includes a discussion of a 2016 book by Michael Tomasello: A natural history of human morality. I think this is relevant background, when one is analysing certain aspects of human morality. A view of human morality, that is genuinely free from unexamined implicit assumptions of non-naturalness, is useful when reading some of the points that I make here on LW. Refs 32 and 46, on agents noticing model misspecification, and on agents interpreting an off switch attempt as an information source, is relevant background to some other topics.
My current research focus is on analysing alignment targets. This is what I post about on LW. I don’t see this as a purely academic curiosity, but instead see it as an issue with important real world implications.
I think that AI is, genuinely, dangerous. The specific AI danger that I am trying to mitigate, is the scenario, where someone successfully hits an alignment target, resulting in an outcome, that is far, far, worse than extinction. I think that this danger, from successfully hitting the wrong alignment target, is severely neglected. I think that this neglect is a genuine problem, and I hope to do something about this neglect, by posting on LW. In other words: since the type of danger that I focus on is distinct, from dangers coming from aiming failures, dealing with it requires dedicated effort (mitigating the types of dangers that I focus on requires a specific type of insights. This class of insights are not useful for dealing with aiming failures. Thus, such insights are unlikely to be found when investigating dangers related to aiming failures).
In yet other words: The danger that I am focused on is importantly distinct from other types of AI dangers. And it requires a dedicated research focus, on issues that are relevant for this specific danger. My research might of course be rendered pointless, by someone accidentally creating an AI, that has no intrinsic interest in humans at all. But I do not think that this is the only outcome worth thinking about. I happen to think that most academic AI researchers, the heads of leading tech companies, the general public, etc, etc, underestimate the probability that a misaligned, uncaring, AI, will kill everyone. If someone manages to trigger an Intelligence Explosion, using current methods, then I don’t expect them to hit the alignment target, that they are aiming for. But I don’t think that this is the only plausible path to a powerful AI. It may be the default path, but many things are uncertain, including the actions of people (for example: Covid illustrated that the set of politically realistic policies, can change dramatically and quickly. And recent AI debate has illustrated that the set of things taken seriously in public debate, can also change dramatically and quickly. These are just two examples of the many sources of uncertainty that I see). In other words: I simply do not think that it is possible, to confidently rule out the possibility, that an alignment target will, eventually, be successfully hit. I further think, that essentially everyone, dramatically underestimate, the dangers associated with successfully hitting the wrong alignment target. My proposed way of reducing this danger, is to analyse alignment targets. My specific focus, in on trying to find features that are necessary (but obviously not sufficient) for an alignment target to be safe for a human individual. Finding such a feature can stop a future AI project (possibly aiming at a currently unknown alignment target) at the idea stage. In scenarios where the results of this type of work has a chance to impact the outcome, I expect that there will be time to pursue this research. However, I have no particular reason to think that there will be, enough, time. If you are also interested in this type of work, then please feel free to send me an email.
Email: thomascederborgsemail at gmail dot com
There is a serious issue with your proposed solution to problem 13. Using a random dictator policy as a negotiation baseline is not suitable for the situation, where billions of humans are negotiating about the actions of a clever and powerful AI. One problem with using this solution, in this contexts, is that some people have strong commitments to moral imperatives, along the lines of ``heretics deserve eternal torture in hell″. The combination of these types of sentiments, and a powerful and clever AI (that would be very good at thinking up effective ways of hurting heretics), leads to serious problems when one uses this negotiation baseline. A tiny number of people with sentiments along these lines, can completely dominate the outcome.
Consider a tiny number of fanatics with this type of morality. They consider everyone else to be heretics, and they would like the AI to hurt all heretics as much as possible. Since a powerful and clever AI would be very good at hurting a human individual, this tiny number of fanatics, can completely dominate negotiations. People that would be hurt as much as possible (by a clever and powerful AI), in a scenario where one of the fanatics are selected as dictator, can be forced to agree to very unpleasant negotiated positions, if one uses this negotiation baseline (since agreeing to such an unpleasant outcome, can be the only way to convince a group of fanatics, to agree to not ask the AI to hurt heretics, as much as possible, in the event that a fanatic is selected as dictator).
This post, explore these issues in the context of the most recently published version of CEV: Parliamentarian CEV (PCEV). PCEV has a random dictator negotiation baseline. The post shows that PCEV results in an outcome massively worse than extinction (if PCEV is successfully implemented, and pointed at billions of humans).
Another way to look at this, is to note that the concept of ``fair Pareto improvements″ has counterintuitive implications, when the question is about AI goals, and some of the people involved, has this type of morality. The concept was not designed with this aspect of morality in mind. And it was not designed to apply to negotiations about the actions of a clever and powerful AI. So, it should not be very surprising, to discover that the concept has counterintuitive implications, when used in this novel context. If some change in the world improves the lives of heretics, then this is making the world worse, from the perspective of those people, that would ask an AI to hurt all heretics as much as possible. For example: reducing the excruciating pain of a heretic, in a way that does not affect anyone else in any way, is not a ``fair Pareto improvement″, in this context. If every person is seen as a heretic by at least one group of fanatics, then the concept of ``fair Pareto improvements″ has some very counterintuitive implications, when it is used in this context.
Yet another way of looking at this, is to take the perspective of human individual Steve, who will have no special influence over an AI project. In the case of an AI, that is describable as doing what a group wants, Steve has a serious problem (and this problem is present, regardless of the details of the specific Group AI proposal). From Steve’s perspective, the core problem, is that an arbitrarily defined abstract entity, will adopt preferences, that is about Steve. But, if this is any version of CEV (or any other Group AI), directed at a large group, then Steve has had no meaningful influence, regarding the adoption of those preferences, that refer to Steve. Just like every other decision, the decision of what Steve-preferences the AI will adopt, is determined by the outcome of an arbitrarily defined mapping, that maps large sets of human individuals, into the space of entities that can be said to want things. Different sets of definitions, lead to completely different such ``Group entities″. These entities all want completely different things (changing one detail can for example change which tiny group of fanatics, will end up dominating the AI in question). Since the choice of entity is arbitrary, there is no way for an AI to figure out that the mapping ``is wrong″ (regardless of how smart this AI is). And since the AI is doing what the resulting entity wants, the AI has no reason to object, when that entity wants the AI to hurt an individual. Since Steve does not have any meaningful influence, regarding the adoption of those preferences, that refer to Steve, there is no reason for him to think that such an AI will want to help him, as opposed to want to hurt him. Combined with the vulnerability of a human individual, to a clever AI that tries to hurt that individual as much as possible, this means that any group AI would be worse than extinction, in expectation.
Discovering that doing what a group wants, is bad for human individuals in expectation, should not be particularly surprising. Groups and individuals are completely different types of things. So, this should be no more surprising, than discovering that any reasonable way of extrapolating Dave, will lead to the death of every single one of Dave’s cells. Doing what one type of thing wants, might be bad for a completely different type of thing. And aspects of human morality, along the lines of ``heretics deserve eternal torture in hell″ shows up throughout human history. It is found across cultures, and religions, and continents, and time periods. So, if an AI project is aiming for an alignment target, that is describable as ``doing what a group wants″, then there is really no reason for Steve to think, that the result of a successful project, would want to help him, as opposed to want to hurt him. And given the large ability of an AI to hurt a human individual, the success of such a project would be massively worse than extinction (in expectation).
The core problem, from the perspective of Steve, is that Steve has no control over the adoption of those preferences, that refer to Steve. One can give each person influence over this decision, without giving anyone any preferential treatment (see for example MPCEV in the post about PCEV, mentioned above). Giving each person such influence, does not introduce contradictions, because this influence is defined in ``AI preference adoption space″, not in any form of outcome space. This can be formulated as an alignment target feature that is necessary, but not sufficient, for safety. Let’s refer to this feature as the: Self Preference Adoption Decision Influence (SPADI) feature. (MPCEV is basically what happens, if one adds the SPADI feature to PCEV. Adding the SPADI feature to PCEV, solves the issue, illustrated by that thought experiment)
The SPADI feature is obviously very underspecified. There will be lots of border cases whose classification will be arbitrary. But there still exists many cases, where it is in fact clear, that a given alignment target, does not have the SPADI feature. Since the SPADI feature is necessary, but not sufficient, these clear negatives are actually the most informative cases. In particular, if an AI project is aiming for an alignment target, that clearly does not have the SPADI feature. Then the success of this AI project, would be worse than extinction, in expectation (from the perspective of a human individual, that is not given any special influence over the AI project). While there are many border cases, regarding what alignment targets could be described as having the SPADI feature, CEV is an example of a clear negative (in other words: there exists no reasonable set of definitions, according to which there exists a version of CEV, that has the SPADI feature). This is because building an AI that is describable as ``doing what a group wants″, is inherent in the core concept, of building an AI, that is describable as: ``implementing the Coherent Extrapolated Volition of Humanity″.
In other words: the field of alignment target analysis is essentially an open research question. This question is also (i): very unintuitive, (ii): very under explored, and (iii): very dangerous to get wrong. If one is focusing on necessary, but not sufficient, alignment target features. Then it is possible to mitigate dangers related to someone successfully hitting a bad alignment target, even if one does not have any idea of what it would mean, for an alignment target to be a good alignment target. This comment outlines a proposed research effort, aimed at mitigating this type of risk.
These ideas also have implications for the Membrane concept, as discussed here and here.
(It is worth noting explicitly that the problem is not strongly connected to the specific aspect of human morality discussed in the present comment (the ``heretics deserve eternal torture in hell″ aspect). The problem is about the lack of meaningful influence, regarding the adoption of self referring preferences. In other words, it is about the lack of the SPADI feature. It just happens to be the case, that this particular aspect of human morality is both (i): ubiquitous throughout human history, and also (ii): well suited for constructing thought experiments, that illustrates the dangers of alignment target proposals, that lack the SPADI feature. If this aspect of human morality disappeared tomorrow, the basic situation would not change (the illustrative thought experiments would change. But the underlying problem would remain. And the SPADI feature would still be necessary for safety).)