You might want to know that I took a look through the site, and was curious, but I just closed the page the moment the “Calculate your contribution” form refused to show me the pricing options unless I gave it an email address.
bogdanb
Karma: 1,235
I’m not sure I understand your weighting argument. Some capabilities are “convergently instrumental” because they are useful for achieving a lot of purposes. I agree that AIs construction techniques will target obtaining such capabilities, precisely because they are useful.
But if you gain a certain convergently instrumental capability, it then automatically allows you to do a lot of random stuff. That’s what the words mean. And most of that random stuff will not be safe.
I don’t get what the difference is between “the AI will get convergently instrumental capabilities, and we’ll point those at AI alignment” and “the AI will get very powerful and we’ll just ask it to be aligned”, other than a bit of technical jargon.
As soon as the AI it gets sufficiently powerful [convergently instrumental capabilities], it is already dangerous. You need to point it precisely at a safe target in outcomes-space or you’re in trouble. Just vaguely pointing it “towards AI alignment” is almost certainly not enough; specifying that outcome safely is the problem we started with.
(And you still have the problem that while it’s working on that someone else can point it at something much worse.)
Exactly. You can’t generalize from “natural” examples to adversarial examples. If someone is trying hard to lie to you about something, verifying what they say can very well be harder than finding the truth would have been absent their input, particularly when you don’t know if and what they want to lie about.
I’m not an expert in any of these and I’d welcome correction, but I’d expect verification to be at least as hard as “doing the thing yourself” in cases like espionage, hacking, fraud and corruption.
AI accelerates the timetable for things we know how to point AI at
It also accelerates the timetable for random things that we don’t expect and don’t even try to point the AI at but that just happen to be easier for incrementally-better AI to do.
Since the space of stuff that helps alignment seems much smaller than the space of dangerous things, you’d expect most things the AI randomly accelerates without us pointing it at will be dangerous.
See above. Don’t become a munitions engineer, and, being aware that someone else will take that role, try to prevent anyone from taking that role. (Hint: That last part is very hard.)
The conclusions might change if planet-destroying bombs are necessary for some good reason, or if you have the option of safely leaving the planet and making sure nobody that comes with you will also want to build planet-destroying bombs. (Hint: That last part is still hard.)
For what it’s worth, the grammar and spelling was much better than is usual for even the native English part of the Internet. That’s probably fainter praise than it deserves, I don’t remember actually noticing any such fault, which probably means there are few of them.
The phrasing and wording did sound weird, but I guess that’s at least one reason why you’re writing, so congratulations and I hope you keep it up! I’m quite curious to see where you’ll take it.
Indeed, the only obvious “power” Harry has that is (as far as we know) unique to him is Partial Transfiguration. I’m not sure if Voldie “knows it not”; as someone mentioned last chapter, Harry used it to cut trees when he had his angry outburst in the Forbidden Forest, and in Azkhaban as well. In the first case Voldie was nearby, allegedly to watch out for Harry, but far enough that to be undetectable via their bond, so it’s possible he didn’t see what exact technique Harry used. In Azkhaban as well he was allegedly unconscious.
I can’t tell if he could have deduced the technique only by examining the results. (At least for the forest occasion he could have made time to examine the scene carefully, and I imagine that given the circumstances he’d have been very interested to look into anything unusual Harry seemed to be able to do.)
On the plus side, Harry performed PT by essentially knowing that objects don’t exist; so it could well be possible to transfigure a thin slice of thread of air into something strong enough to cut. For that matter, that “illusion of objects” thing should allow a sort of “reverse-Partial” transfiguration, i.e. transfigure (parts of) many objects into a single thing. Sort of like what he did to the troll’s head, but applied simultaneously to a slice of air, wands, and Death Eaters. Dumbledore explicitly considers it as a candidate against Voldemort (hint, Minerva remembers Dumbledore using transfiguration in combat). And, interestingly, it’s a wordless spell (I’m not even sure if Harry can cast anything* else wordlessly), and Harry wouldn’t need to raise his wand, or even move at all, to cast it on air (or on the time-space continuum, or world wave-function, whatever).
On the minus side, I’m not sure if he could do it fast enough to kill the Death Eaters before he’s stopped. He did get lots of transfiguration training, and using it in anger in the forest suggests he can do it pretty fast, but he is watched, and IIRC transfiguration is not instantaneous. He probably can’t cast it on Voldie nor on his wand, though he might be able to destroy the gun. And Voldemort can certainly find lots of ways to kill him without magic or touching him directly; hell, he probably knows kung fu and such. And even if Harry managed to kill this body, he’d have to find a way to get rid of the Horcruxes. (I still don’t understand exactly what the deal is with those. Would breaking the Resurrection Stone help?)
Well, we only know that Harry feels doom when near Q and/or his magic, and that in one case in Azkhaban something weird happened when Harry’s Patronus interacted with what appeared to be an Avada Kedavra bolt, and that Q appears to avoid touching Harry.
Normally I’d say that faking the doom sensations for a year, and faking being incapacitated while trying to break someone out of Azkhaban, would be too complicated. But in this case...
Both good points, thank you.
Thank you, that was very interesting!
I sort of get your point, but I’m curious: can you imagine learning (with thought-experiment certainty) that there is actually no reality at all, in the sense that no matter where you live, it’s simulated by some “parent reality” (which in turn is simulated, etc., ad infinitum)? Would that change your preference?
most “earthlike” planets in habitable zones around sunlike stars are on average 1.8 Billion years older than the Earth
How do you know? (Not rhethorical, I have no idea and I’m curious.)
If the final goal is of local scope, energy acquisition from out-of-system seems to be mostly irrelevant, considering the delays of space travel and the fast time-scales a strong AI seems likely to operate at. (That is, assuming no FTL and the like.)
Do you have any plausible scenario in mind where an AI would be powerful enough to colonize the universe, but do it because it needs energy for doing something inside its system of origin?
I might see one perhaps extending to a few neighboring systems in a very dense cluster for some strange reason, but I can’t imagine likely final goals (again, for its birth star-system) that it would need to spend hundreds of millenia even to take over a single galaxy, let alone leave it. (Which is of course no proof there isn’t; my question above wasn’t rhethorical.)
I can imagine unlikely accidents causing some sort of papercliper-scenario, and maybe vanishingly rare cases where two or more AIs manage to fight each other over long periods of time, but it’s not obvious to me why this class of scenarios should be assigned a lot of probability mass in aggregate.
Honestly, I can’t really find anything significant in this comment I disagree with.
It’s a bit like opening a thread arguing that the Spanish inquisition was right for torturing nonbelievers because they they acted under the assumption that they could save souls from eternal damnation by doing so.
But the OP didn’t argue in support of torturing people, as far as I can tell. In the terms of your analogy, my reading was of the OP was a bit like:
“Hey, if the Spanish Inquisition came to you and offered the following two options, would you pick either of them, or refuse both? The options are (1) you’re excommunicated, then you get all the cake you want for a week, then you forget about it, or (2) you’re sanctified, then you’re tortured for a week, then you forget about it. Option (3) means nothing happens, they just leave.”
Which sounds completely different to my ears.
Sure, but then why do you expect memory and experience would also behave in a common sense manner? (At least, that’s what I think you did in your first comment.)
I interpreted the OP as “I’m confused about memory and experience; let’s try a thought experiment about a very uncommon situation just to see what we think it would happen”. And your first comment reads to me as “you picked a bad thought experiment, because you’re not describing a common situation”. Which seems to completely miss the point, the whole purpose of the thought experiment was to investigate the consequences of something very distinct from situations where “common sense” has real experience to rely on.
The part about torturing children I don’t even get at all. Wondering about something seems to me almost the opposite of the philosophy of “doing something because you think you know the answer”. Should we never do thought experiments, because someone might act on mistaken assumptions about those ideas? Not thinking about something before doing it sounds to me like exactly the opposite of the correct strategy.
Once AI is developed, it could “easily” colonise the universe.
I was wondering about that. I agree with the could, but is there a discussion of how likely it is that it would decide to do that?
Let’s take it as a given that successful development of FAI will eventually lead to lots of colonization. But what about non-FAI? It seems like the most “common” cases of UFAI are mistakes in trying to create an FAI. (In a species with similar psychology to ours, a contender might also be mistakes trying to create military AI, and intentional creation by “destroy the world” extremists or something.)
But if someone is trying to create an FAI, and there is an accident with early prototypes, it seems likely that most of those prototypes would be programmed with only planet-local goals. Similarly, it doesn’t seem likely that intentionally-created weapon-AI would be programmed to care about what happens outside the solar system, unless it’s created by a civilization that already does, or is at least attempting, interstellar travel. Creators that care about safety will probably try to limit the focus, even imperfectly, both to make reasoning easier and to limit damage, and weapons-manufacturers will try to limit the focus for efficiency.
Now, I realize that a badly done AI could decide to colonize the universe even if its creators didn’t program it for that initially, and that simple goals can have that as an unforeseen consequence (like the prototypical paperclip manufacturer). But have we any discussion of how likely that is in a realistic setting? Perhaps the filter is that the vast majority of AIs limit themselves to their original solar system.
The problem with that is that life on Earth appeared about 4 billion years ago, while the Milky Way is more than 13 billion years old. If life were somewhat common, we wouldn’t expect to be the first, because there was time for it to evolve several times in succession, and it had lots of solar systems where it could have done it.
A possible answer could be that there was a very strong early filter during the first part of the Milky Way’s existence, and that filter lessened in intensity in the last few billion years.
The only examples I can think of are elemental abundance (perhaps in a young galaxy there are much fewer systems with diverse enough chemical compositions) and supernova frequency (perhaps a young galaxy is sterilized by frequent and large supernovas much more often than an older one’s). But AFAIK both of those variations can be calculated well enough for a Fermi estimate from what we know, so I’d expect someone who knows the subject much better than I would have made that point already if they were plausible answers.
Your rephrasing essentially says that you torture an identical copy of a person for a week.
If you read it carefully, my first rephrasing actually says that you torture the original person for a week, and then you (almost) perfectly erase their memories (and physical changes) during that week.
This is not changing the nature of the thought experiment in the OP; it is exactly the same experiment, plus a hypothetical example of how it could be achieved technically, because you implied that the experiment in the OP is impossible to achieve and thus ill-posed.
Or, at least, that’s how I interpreted “Of course I’m fighting the hypothetical thought experiment. I think the notion of experience without being affected doesn’t make any sense.” I just gave an example of how one can experience something and not be affected. It was a somewhat extreme example, but it seems appropriate when Omega is involved.
And yet there is a moment where the translator emits a very long sentence in a way that suggests that the translation suite is able to paraphrase compact concepts we don’t have a word for: