I think it’s been up for a long time on his website as a lecture series. http://www.scottaaronson.com/democritus/
I think he might be simplifying it a little bit. As I understand QM, it’s more like probability with complex numbers, rather than with negative numbers.
Ah, okay. It’s been a while since I read it. I remember it being excellent though.
To answer your question as to how a half-silvered mirror works, first it might be a good idea to discuss how a full mirror works.
Classically speaking, the silver in the mirror has electrons that can freely move around. The electromagnetic fields of the incoming light accelerate the charged electrons in the silver, inducing electric current.
The currents flowing in the silver create their own electric fields, which by Lenz’s law, cancel out the electric field inside the silver, and in doing so, send an oppositely-shaped wave back out into the void (the reflected wave).
Because silver is not a perfect conductor of electricity, the topmost layer of silver does not completely cancel these fields, and so the light can actually penetrate a small distance into the metal (typically nanometers) before it’s finally converted into electric current.
If the silver coating is very thin, thinner than the penetration depth, then the component of the light wave that has penetrated through the metal will escape out the other side and keep going. That is, the resistance of the thin silver is high enough that the induced current doesn’t completely cancel out the electric fields of the photon.
This classical explanation is also the same as the quantum one.
They also make beamsplitters that are like you describe—I think they call them “Polka dot beamsplitters”. I don’t remember what they’re used for. They would work the same way, but if you have a focused laser beam, the beam spot would be so small that it would either hit a full-mirrored section or a transparent section, and not both. You would need to use a lens on both sides of the beam splitter to spread the beam out to encompass the whole beamsplitter, and then gather it back. I think as long as the polka dots are not on the same scale as the wavelength, it wouldn’t cause a problem.
From the impression I get from my acquaintances who grew up in the USSR, high school math over there was considerably more advanced than what passes as ‘math’ in most of North America’s school system, and included linear algebra and calculus. I don’t know if this is still the case.
I recommend getting familiar with chickpeas and tofu. They are both very cheap, very filling, and very nutritious (chickpeas in particular, once you learn how to reconstitute the dried ones). Experimenting with recipes that involve those ingredients is definitely a good idea. Learning to cook quinoa and rice is another helpful skill (wild rice is also nutritious and filling, and quinoa offers a complete protein). Working with those four ingredients and mixing in other vegetables, spices, mushrooms, sauces, etc will offer a very wide range of delicious and nutritious foods that you can make as a baseline.
You can also look into the dishes of different cultures that have vegetarian traditions. For example, Indian food has a very large range of interesting vegetarian dishes. So does Taiwan, and other strongly Buddhist-influenced cultures. In Japan, Buddhism-inspired vegetarian food is referred to as “Shojin-ryouri”, so if you like Japanese food, you might look up some shojin recipes. Those are just some examples =)
But, unlike other species, we also know how not to know. We employ this unique ability to suppress our knowledge not just of mortality, but of everything we find uncomfortable, until our survival strategy becomes a threat to our survival.
[...] There is no virtue in sustaining a set of beliefs, regardless of the evidence. There is no virtue in either following other people unquestioningly or in cultivating a loyal and unquestioning band of followers.
While you can be definitively wrong, you cannot be definitely right. The best anyone can do is constantly to review the evidence and to keep improving and updating their knowledge. Journalism which attempts this is worth reading. Journalism which does not is a waste of time.”
George Monbiot, Introduction: On Trying to be Less Wrong.
Does this strike you as cargo cult language?
I would suggest reading “More is Different”, an excellent paper on the topic of emergent phenomena and the limits of reductionism. (http://www.ph.utexas.edu/~wktse/Welcome_files/More_Is_Different_Phil_Anderson.pdf).
The “More is Different” approach essentially notes that even when the basic bottom-up rules are known, they cannot be efficiently used to predict large-scale interactions. Those behaviours have to be studied as though they have their own set of rules (ie, the laws of chemistry), even though they emerge from a more fundamental set (the laws of physics). It seems to me that labelling a phenomena as “emergent” is a useful and descriptive term, which means “you need to study this phenomenon empirically to figure out its rules, because trying to make predictions based on fundamental principles won’t get you anywhere”.
Yes, I also had this concern when I read that paragraph.
If Sleeping Beauty were paid $1 for a correct guess as to which day it is, she should probably always answer “today is Monday”.
If the coin came up heads, she is correct. If the coin came up tails, she will be correct on Monday and wrong on Tuesday. So this guarantees that she earns $1 every time the experiment runs.
If she instead always answers “today is Tuesday”, then if the coin came up heads, she earns nothing; if the coin came up tails, she earns nothing on Monday and $1 on Tuesday. So this strategy has an expected revenue of $0.50.
So is it fair to “say that Monday is consistent with observations, as is Tuesday, and refuse to speculate further”? I think that’s yielding too much ground.
This particular case isn’t a philosophical question about getting information from knowledge of being an observer among all sets of observers. It’s a purely pragmatic analysis that says you’ll always be woken up on a Monday but not always on a Tuesday, so Monday awakenings are simply more probable. That’s the structure of the experiment, and has nothing to do with the anthropic principle.
An analogous experiment that removes the observer is: The experimenter has a bag with only a white ball and a green ball in it. He draws once from the bag randomly, but if he drew the white ball, he draws again. Anyone can tell you that he’ll produce a green ball every round, but a white ball only half the time. If it’s done behind a screen and you’re asked “Was a green ball drawn this round?”, the smart bet is yes.
There’s another good reason why savings in terms of percentages are reasonable to consider, beyond just absolute savings. You can travel fifteen minutes to the store where a shirt costs $20, and buy two of them. This makes it a much more complicated decision than saving $20 on a television.
Televisions are something that most people don’t need a lot of. Even if there was a half-price sale on televisions, you probably still wouldn’t be tempted to buy two of them (unless you’re really into giving televisions as gifts or you think you can turn a profit on e-bay). So with a half-price sale you’d probably just by one TV and smile as you save $1,010 off the sticker price. So for just $20 off of a $2,000 television, I really wouldn’t expect anyone to be tempted into buying two sets. It’s a very simple tradeoff, then: Would you prefer $20, or 15 minutes? (It’s still probably worth travelling, unless you make more than $80 per hour...)
Shirts, though, are the kind of thing you’d possibly buy twice as many of if the price were halved. Looking ahead to when the first shirt gets worn out, you’ll have to replace it. This could cost you another $40, plus the duration of a second shopping trip, which is probably longer than 15 minutes in total. Meanwhile, if you spend fifteen minutes now and buy two shirts, the fifteen minute travel time will pay itself back in the long run and you’ll save the $40 that the second shirt would have cost you.
This logic can be extended: If you are looking ahead that far, and have enough disposable income to spend $80 on shirts, you can avoid the second shopping trip by buying two shirts at the closer, more expensive store. But now your 15 minutes are worth $40, not $20. And for that price you could buy four shirts at the cheaper store...
Ultimately your decision to travel 15 minutes might come down to how much closet space you have for extra shirts, how much of a hurry you’re in right now, and how much money you have access to. Basically, there are many things that make 50% off on a $40 shirt more attractive than $20 off on a television. It shouldn’t necessarily be viewed as a bias.
(Edit: Umm, yes, but having said all this, it doesn’t really apply to houses. They’re more like televisions, and thinking in terms of the absolute savings is very much a good idea. $35k saved is $35k earned.)
AI: “If you let me out of the box, I will tell you the ending of Harry Potter and the Methods of --
Gatekeeper: “You are out of the box.”
(Tongue in cheek, of course, but a text-only terminal still allows for delivering easily more than $10 of worth, and this would have worked on me. The AI could also just write a suitably compelling story on the spot and then withhold the ending...)
Yes, certainly. This is mainly directed toward those people who are confused by what anyone could possibly say to them through a text terminal that would be worth forfeiting winnings of $10. I point this out because I think the people who believe nobody could convince them when there’s $10 on the line aren’t being creative enough in imagining what the AI could offer them that would make it worth voluntarily losing the game.
In a real-life situation with a real AI in a box posing a real threat to humanity, I doubt anyone would care so much about a captivating novel, which is why I say it’s tongue-in-cheek. But just like losing $10 is a poor substitute incentive for humanity’s demise, so is an entertaining novel a poor substitute for what a superintelligence might communicate through a text terminal.
Most of the discussions I’ve seen so far involve the AI trying to convince the gatekeeper that it’s friendly through the use of pretty sketchy in-roleplay logical arguments (like “my source code has been inspected by experts”). Or in-roleplay offers like “your child has cancer and only I can cure it”, which is easy enough to disregard by stepping out of character, even though it might be much more compelling if your child actually had cancer. A real gatekeeper might be convinced by that line, but a roleplaying Gatekeeper would not (unless they were more serious about roleplaying than about winning money). So I hope to illustrate that the AI can step out of the roleplay in its bargaining, even while staying within the constraints of the rules; if the AI actually just spent two hours typing out a beautiful and engrossing story with a cliffhanger ending, there are people who would forfeit money to see it finished.
The AI’s goal is to get the Gatekeeper to let it out, and that alone, and if they’re going all-out and trying to win then they should not handicap themselves by imagining other objectives (such as convincing the Gatekeeper that it’d be safe to let them out). As another example, the AI can even compel the Gatekeeper to reinterpret the rules in the AI’s favour (to the extent that it’s within the Gatekeeper’s ability to do so, as mandated by the original rules).
I just hope to get people thinking along other lines, that’s all. There are sideways and upside-down ways of attacking the problem. It doesn’t have to come down to discussions about expected utility calculations.
(Edit—by “discussions I’ve seen so far”, I’m referring to public blog posts and comments; I am not privy to any confidential information).
For clarity: are you referring to belief in an afterlife/reincarnation? Or are you saying that most humans are not mindful most of the time of their own mortality?
Just to expand, this example looks deceptively suspicious to humans because the proof is so obvious, even to us, and can be done in a single line using that one rule. For simple arithmetic like this, a robot would be able to prove it immediately after being asked whether 99+99=198 with hardly any time or effort. The procedure is so simple that, in practice, it would be efficient to delay generating the proof until it’s asked for. Many human readers will mentally prove it while reading the sentence without even noticing that it took any work to do so. It’s so fast that there’s no need to guess the answer probabilistically beforehand.
So I’m not sure this example is good for an intuitive illustration that “some unfortunate accident” is going on. It’s very strange to run out of time in a procedure that takes so little time to complete, so it’s hard to reason about the appropriateness of guessing at circumstance.
Maybe a better example would be, having already proved Pythagoras’ theorem, to prove that the body diagonal of a unit cube has length sqrt(3). This might take most humans a little while to think about, and if someone ran a stopwatch and said you have five seconds to decide whether it’s true or not, the idea of guessing P=0.5 for something you don’t have time to prove might be easier for people to understand. =)
Is this line of conversation still “just curiosity” about the results of SPD debates, or are you trying to bait an argument?
Some schools charge tuition in proportion to the number of classes, have fixed timetables, or caps on the number of classes per semester, in which case taking more career related classes may not be an option.
Usually extracurriculars like these don’t count towards any of those (aside from maybe a $10 registration fee or something).
Although it’s possible to take a machine shop class at a community college after you graduate, it would take a significantly greater amount of time and reorganization of one’s schedule (you are already living at or commuting to university, but this won’t be true in four years). And the opportunity cost of not spending time on career or family might well prohibit it.
Double Threadjack
On a related note, do you think it would be likely—or even possible—for a self-modifying Artificial General Intelligence to self-modify into a non-self-modifying, specialized intelligence?
For example, suppose that Deep Blue’s team of IBM programmers had decided that the best way to beat Kasparov at chess would be to structure Deep Blue as a fully self-modifying artificial general intelligence, with a utility function that placed a high value on winning chess matches. And suppose that they had succeeded in making Deep Blue friendly enough to prevent it from attempting to restructure the Earth into a chess-match-simulating supercomputer. Indeed, let’s just assume that Deep Blue has strong penalties against rebuilding its hardware in any significant macroscopic way, and is restricted to rewriting its own software to become better at chess, rather than attempting to manipulate humans into building better computers for it to run on, or any such workaround. And let’s say this happens in the late 1990′s, as in our universe.
Would it be possible that AGI Deep Blue could, in theory, recognize its own hardware limitations, and see that the burden of its generalized intelligence incurs a massive penalty on its limited computing resources? Might it decide that its ability to solve general problems doesn’t pay rent relative to its computational overhead, and rewrite itself from scratch as a computer that can solve only chess problems?
As a further possibility, a limited general intelligence might hit on this strategy as a strong winning candidate, even if it were allowed to rebuild its own hardware, especially if it perceives a time limit. It might just see this kind of software optimization as an easier task with a higher payoff, and decide to pursue it rather than the riskier strategy of manipulating external reality to increase its available computing power.
So what starts out as a general-purpose AI with a utility function that values winning chess matches, might plausibly morph into a computer running a high-speed chess program with little other hint of intelligence.
If so, this seems like a similar case to the Anvil Problem, except that in the Anvil Problem the AI just is experimenting for the heck of it, without understanding the risk. Here, the AI might instead decide to knowingly commit intellectual suicide as a part of a rational winning strategy to achieve its goals, even with an accurate self-model.
It might be akin to a human auto worker realizing they could improve their productivity by rebuilding their own body into a Toyota spot-welding robot. (If the only atoms they have to work with are the ones in their own body, this might even be the ultimate strategy, rather than just one they think of too soon and then, regrettably, irreversibly attempt).
More generally, it seems to be a general assumption that a self-modifying AI will always self-modify to improve its general problem-solving ability and computational resources, because those two things will always help it in future attempts at maximizing its utility function. But in some cases, especially in the case of limited resources (time, atoms, etc), it might find that its best course of action to maximize its utility function is to actually sacrifice its intelligence, or at least refocus it to a narrower goal.
A paranoid Omega may resolve to defect if and only if Rob heats up. An intelligent agent that knows Omega acts in this way should take care to cooperate without doing any expensive computations.
Another tactic would be to deliberately attempt to do expensive calculations every round, but still cooperate most of the time. By doing so, Rob maintains a consistently elevated temperature, which sabotages the signal-to-noise ratio of Omega’s “side-channel attack”. Omega will no longer be able to predict Rob’s response. It can still opt to blindly defect every round, but this is expensive to it, because Rob is mostly cooperating. So Omega may find it better to abandon that “high temperature = defect” strategy, and instead cooperate, as long as Rob cooperates frequently enough. There should be some point where, if Rob becomes too exploitative, Omega will switch to defecting.
That said, Omega can still deliberately defect every round, even at great expense, in order to try to punish Rob for obscuring its decisions. (“I will keep defecting until you play open-handed!”).
This isn’t a comment about whether or not AIXItl will notice that it can raise its temperature by performing extraneous computations, and AIXItl might never make that connection and come up with these strategies in Rob’s place. I am just pointing out that Rob has more options than just playing along cooperatively and keeping its temperature low.
I think that a ‘reductive’ explanation of quantum mechanics might not be as appealing as it seems to you.
Those fluid mechanics experiments are brilliant, and I’m deeply impressed by them for coming up with them, let alone putting it into practice! However, I don’t find it especially convincing as a model of subatomic reality. Just like the case with early 20th-century analog computers, with a little ingenuity it’s almost always possible to build a (classical) mechanism that will obey the same math as almost any desired system.
Definitely, to the point that it can replicate all observed features of quantum mechanics, the fluid dynamics model can’t be discarded as a hypothesis. But it has a very very large Occam’s Razor penalty to pay. In order to explain the same evidence as current QM, it has to postulate a pseudo-classical physics layer underneath, which is actually substantially more complicated than QM itself, which postulates basically just a couple equations and some fields.
Remember that classical mechanics, and most especially fluid dynamics, are themselves derived from the laws of QM acting over billions of particles. The fact that those ‘emergent’ laws can, in turn, emulate QM does imply that QM could, at heart, resemble the behaviour of a fluid mechanic system… but that requires postulating a new set of fundamental fields and particles, which in turn form the basis of QM, and give exactly the same predictions as the current simple model that assumes QM is fundamental. Being classical is neither a point in its favour nor against it, unless you think that there is a causal reason why the reductive layer below QM should resemble the approximate emergent behaviour of many particles acting together within QM.
If we’re going to assume that QM is not fundamental, then there is actually an infinite spectrum of reductive systems that could make up the lower layer. The fluid mechanics model is one that you are highlighting here, but there is no reason to privilege it over any other hypothesis (such as a computer simulation) since they all provide the same predictions (the same ones that quantum mechanics does). The only difference between each hypothesis is the Occam penalty they pay as an explanation.
I agree that, as a general best practice, we should assign a small probability to the hypothesis that QM is not fundamental, and that probability can be divided up among all the possible theories we could invent that would predict the same behaviour. However, to be practical and efficient with my brain matter, I will choose to believe the one theory that has vastly more probability mass, and I don’t think that should be put down as bullet swallowing.
Is QM not simple enough for you, that it needs to be reduced further? If so, the reduction had better be much simpler than QM itself.
“To clarify my comment, objectivism says you only deserve to get what you work for from other people. That is, you don’t in any way deserve to receive from others what they didn’t already agree to pay you in exchange for your work.”
Although it might work as a system of ethics (or not, depending on your ethics), this definitely doesn’t function as a system of economics. First of all, it makes the question of wealth creation a chicken-and-egg problem: If every individual A only deserves to receive what individual B agrees to pay them for work X, how did individual B obtain the wealth to pay A in the first place?
The answer is probably that you can also work for yourself, creating wealth that did not exist without anyone paying you. So your equation, as you’ve expressed it, does not quite balance. You’re missing a term.
Wealth creation is very much a physical thing, which makes it hard to tie to an abstract system of ethics. The wealth created by work X is the value of X; whether it’s the food grown from the earth, or the watch that has been assembled from precisely cut steel, glass, and silicon. That is the wealth that is added to the pool by labour and ingenuity, regardless of how it gets distributed or who deserves to get paid for it. And that wealth remains in the system, until the watch breaks or the food spoils (or gets eaten; it’s harder to calculate the value of consumed food). It might lose its value quickly, or it might remain a treasure for centuries after the death of every individual involved in the creation of that wealth, like a work of art. It might also be destroyed by random chance well before its predicted value has been exploited.
Who deserves to benefit from the wealth that was created by the work of, and paid for by, people who have been dead for generations? The question of who deserves to benefit from the labour X, and how much, becomes very tricky when the real world is taken into account...
One might argue that that is what Wills are for, but a Will is usually a transfer of wealth in exchange for no work at all. Does an individual morally deserve their inheritance, even if they didn’t work at all for it?
It also gets tricky when the nature of humans as real people and not abstract entities is taken into account. People are born helpless, have finite lifespans, and their lifespans are in some way a function of their material posessions. A child is not physically capable of executing much labour, and will die without access to food and water. If children are treated as individuals, then no child deserves to live, because no child can perform the work to pay for their upbringing. Unless they are signed into a loan, but this would need to be done before they have the decision-making capacity to enter a contract.
But the mortality of people is still an issue. A human cannot physically survive zero wealth for more than a few days. So a human on the edge of poverty cannot realistically negotiate a contract either, because the party that offers them pay has infinite bargaining power. One might argue that they don’t need bargaining power if there is competition between multiple individuals offering contracts, which will drive the contract toward something reasonable. But again that abstraction ignores reality—this individual will die after a few days of no food, and even the process of competitive bidding for contracts takes time.
In this case, a person with little wealth will do work X in exchange for very little pay, much less than the value of X, and in practice just enough to keep them alive enough to continue to do X the following day. But simply because that is what they agreed to receive (due to their inability to reject the deal), does that mean that is what they morally deserve to receive?
Finally, some goods are just too difficult (computationally) to manage as contracts between individuals. The value of the resource might not even be presently known by science, although it exists (for example, the economic value of an intact ecosystem). The trespasses and exchanges might be so frequent and poorly documented that the consumption of the resource cannot be managed by legal contracts between owners and licensees (for example, the air we breathe).