I recently got reminded of this post. I’m not sure I agree with it, because I think we have different paradigms for AI alignment—I’m not nearly so concerned with the sort of oversight that relies on looking at the state of the computer. Though I have nothing against the sort of oversight where you write a program to tell you about what’s going on with your model.
Instead, I think that anticipating the effects of QC on AI alignment is a task in prognosticating how ML is going to change if you make quantum computing available. I think the relevant killer app is not going to be Grover’s algorithm, but quantum annealing. So we have to try to think about what kind of ML you could do if you could get a large speedup on optimization on objective functions but were limited to a few hundred to a few thousand bits at a time (assuming that that’s true for near-future quantum computers).
And given those changes, what changes for alignment? Seems like a hard question.
I think the relevant killer app is not going to be Grover’s algorithm, but quantum annealing.
Minor comment here: quantum annealing can be thought of basically, as the equivalent of quantum walks, which are an extension of Grover’s algorithm. I would be very surprised if there was any difference between them.
I do not think we really disagree on this point. I also believe that looking at the state of the computer is not as important as having an understanding of how the program is going to operate and how to shape its incentives.
Maybe this could be better emphasized, but the way I think about this article is showing that even the strongest case for looking at the intersection of quantum computing and AI alignment does not look very promising.
re: How quantum computing will affect ML
I basically agree that the most plausible way QC can affect AI aligment is by providing computational speedups—but I think this mostly changes the timelines rather than violating any specific assumptions in usual AI alignment research.
Relatedly, I am bearish that we will see better than quadratic speedups (ie Grover) - to get better-than-quadratic speedups you need to surpass many challenges that right now it is not clear can be surpassed outside of very contrived problem setup [REF].
In fact I think that the speedups will not even be quadratic because you “lose” the quadratic speedup when parallelizing quantum computing (in the sense that the speedup does not scale quadratically with the number of cores).
I recently got reminded of this post. I’m not sure I agree with it, because I think we have different paradigms for AI alignment—I’m not nearly so concerned with the sort of oversight that relies on looking at the state of the computer. Though I have nothing against the sort of oversight where you write a program to tell you about what’s going on with your model.
Instead, I think that anticipating the effects of QC on AI alignment is a task in prognosticating how ML is going to change if you make quantum computing available. I think the relevant killer app is not going to be Grover’s algorithm, but quantum annealing. So we have to try to think about what kind of ML you could do if you could get a large speedup on optimization on objective functions but were limited to a few hundred to a few thousand bits at a time (assuming that that’s true for near-future quantum computers).
And given those changes, what changes for alignment? Seems like a hard question.
Minor comment here: quantum annealing can be thought of basically, as the equivalent of quantum walks, which are an extension of Grover’s algorithm. I would be very surprised if there was any difference between them.
Other than that, I agree with your comment but also wanted to notice this article: https://journals.aps.org/prxquantum/abstract/10.1103/PRXQuantum.2.010103 which poses severe difficulty to the usefulness of any Grover-like advantage.
re: impotance of oversight
I do not think we really disagree on this point. I also believe that looking at the state of the computer is not as important as having an understanding of how the program is going to operate and how to shape its incentives.
Maybe this could be better emphasized, but the way I think about this article is showing that even the strongest case for looking at the intersection of quantum computing and AI alignment does not look very promising.
re: How quantum computing will affect ML
I basically agree that the most plausible way QC can affect AI aligment is by providing computational speedups—but I think this mostly changes the timelines rather than violating any specific assumptions in usual AI alignment research.
Relatedly, I am bearish that we will see better than quadratic speedups (ie Grover) - to get better-than-quadratic speedups you need to surpass many challenges that right now it is not clear can be surpassed outside of very contrived problem setup [REF].
In fact I think that the speedups will not even be quadratic because you “lose” the quadratic speedup when parallelizing quantum computing (in the sense that the speedup does not scale quadratically with the number of cores).