There are a significant number of subjects that philosophers used to discuss, before science was able to study them, then later science started to cast light on them, eventually identified the correct hypotheses, and the philosophers gradually lost interest in them. So there is and alternative to the above: identify all the alignment-relevant questions that we currently have philosophical hypotheses about but not scientific answers, and develop scientific answers to them. (AI-Assisted Research comes to mind as a strategy to do this sooner — though obviously doing AI-Assisted Research before you’ve solved the alignment problem has some chicken-and-egg issues.)
For example, philosophers spend a significant amount of time discussing the nature of consciousness. We are learning how to build minds, and we may at some point understand neuroscience well enough to be able to do human uploading. At some point, we will have scientific answers to many of the questions about consciousness that we currently only have philosophical ideas about. Those, judging by the past history of these things, may turn out to be much more complex and/or mind-bending that any of the ideas philosophers explored (atomism is a counterexample, but not a recent one — quantum mechanics is pretty strong recent example).
Many alignment-related philosophical issues relate to ethics and morality. The scientific field aimed at addressing this is Evolutionary Moral Psychology. Sadly it’s far easier to come up with plausible hypotheses about the evolutionarily optimum behaviors for social hominds than it is to actually test them, so progress is slow. Doing in-silico simulation experiments might be able to speed this up.
There are a significant number of subjects that philosophers used to discuss, before science was able to study them, then later science started to cast light on them, eventually identified the correct hypotheses, and the philosophers gradually lost interest in them. So there is and alternative to the above: identify all the alignment-relevant questions that we currently have philosophical hypotheses about but not scientific answers, and develop scientific answers to them. (AI-Assisted Research comes to mind as a strategy to do this sooner — though obviously doing AI-Assisted Research before you’ve solved the alignment problem has some chicken-and-egg issues.)
For example, philosophers spend a significant amount of time discussing the nature of consciousness. We are learning how to build minds, and we may at some point understand neuroscience well enough to be able to do human uploading. At some point, we will have scientific answers to many of the questions about consciousness that we currently only have philosophical ideas about. Those, judging by the past history of these things, may turn out to be much more complex and/or mind-bending that any of the ideas philosophers explored (atomism is a counterexample, but not a recent one — quantum mechanics is pretty strong recent example).
Many alignment-related philosophical issues relate to ethics and morality. The scientific field aimed at addressing this is Evolutionary Moral Psychology. Sadly it’s far easier to come up with plausible hypotheses about the evolutionarily optimum behaviors for social hominds than it is to actually test them, so progress is slow. Doing in-silico simulation experiments might be able to speed this up.
[For a more in-depth discussion of the last paragraph, see Grounding Value Learning in Evolutionary Psychology: an Alternative Proposal to CEV.]