If you replace “true” with “accurate,” what does “accurate” mean?
I would have thought that “accurate” means that the distance between the model result and the true result is small, so it contains a notion of truth and a notion of distance.
If you replace “true” with “accurate,” what does “accurate” mean?
I would have thought that “accurate” means that the distance between the model result and the true result is small, so it contains a notion of truth and a notion of distance.
Okay, I just did a deep-dive on the AI alignment problem and the Singularity on Wikipedia, and it will take me a while to digest all of that. My first impression is that it seems like an outlandish thing to worry about, but I am going to think about it more because I can easily imagine the situation reversed.
Among the things I came across was that Eliezer was writing about this in 1996, and predicted
Plug in the numbers for current computing speeds, the current doubling time, and an estimate for the raw processing power of the human brain, and the numbers match in: 2021.
GPT-3 has some tens to hundreds of billion parameters and the human brain has 86 billion neurons, and I know it’s hand-waving because model parameters aren’t equivalent to human neurons, but—not bad! On the other hand, we’re seeing now what this numerical correspondence translates to in real life, and it’s interestingly different than what we I had imagined. AI is passing the Turing test, but the Turing test no longer feels like a hard line in the sand; it doesn’t seem to be testing what it was intended to test.
No one knows how to solve it, and it’s likely only rationalists could.
Understanding what, exactly, human values are would be a first step toward expressing it in AI. I hadn’t expected meta-ethics to get so applied.
...
You know what’s really odd? The word “singularity” appears only 33 times in the Sequences, mostly as “when I attended the Singularity Summit, someone said...” and such, without explanation of what it was. Most of the references were in the autobiographical section, which I didn’t read as deeply as the rest.
This looks a lot like a typical high school/college freshman physics problem, and I guess the moral of the story is that it leads us to think that we should solve it that way. But if you were to work it out,
I think the ball’s rotational energy would be a much smaller number than the gravitational potential energy of falling a few feet. The rotational energy of a solid sphere is , where and are the mass and radius of the ball and is the angular velocity of rotation. Meanwhile, the gravitational potential energy is , where . There are some quantities whose values we don’t know, like , but looking at the set-up, I seriously doubt that rotational energy, or lack thereof because the ball doesn’t stick to the track, is going to matter.
Fun fact: Galileo didn’t drop weights off the Leaning Tower of Pisa; he rolled balls down slopes like this. He completely ignored/didn’t know about rotational energy, and that was an error in his measurements, but it was small enough to not change the final result. He also used his heartbeat as a stopwatch.
I think the biggest effect here is the bendy track. It’s going to absorb a lot (like ) of the energy, and can’t be ignored. Alison uses the questioner’s motives as data (“Calculating the effect of the ramp’s bendiness seems unreasonably difficult and this workshop is only meant to take an hour or so, so let’s forget that.”), which she shouldn’t.
This is both interesting and (I think) an important thing to know about science: plans and strategies are systematic, but discoveries sometimes are and sometimes aren’t. In particle physics, the Omega baryon and Higgs boson were discovered in deliberate hunts, but the muon and J/psi were serendipitous. The ratio might be about half-and-half (depending on how you count particles).
Thinking about this, I have two half-answers, which may be leads as to why sweetener discovery might be discovered by serendipity, even though there are systematic searches for new drugs.
Discovery depends, to a great degree, on your detector, and I don’t think there’s a better detector of sweetness than the ones in our mouths. Presumably, searches through virtual (not synthesized) molecules can be faster, and if the identification algorithm can accurately predict activation of the sweetness receptor, then it could outperform detection by taste only because it’s faster than synthesis. But virtual drug discovery is still an open problem, still under development...
Maybe there are, in nature, only a few sweet molecules, and they were discovered early. Going through the list of artificial sweeteners you mentioned, below are the discovery dates. When were most of the systematic drug searches? Did it cover this timespan, which seems to be in the early and mid-20th century?
Saccharin: 1897
Cyclamate: 1937
Aspartame: 1965
Acesulfame potassium: 1967
Sucralose: 1976
(This suggestion also has an analogy with particle physics: hundreds of particles were discovered in the 1950′s because accelerators had just been invented that could illuminate the strong-force mass range, which has rich phenomenology. At the current frontier, though, there are very few particles.)
Other comments in the comments section that sound quite likely to me are: (1) perhaps the very sweet compounds could be smelled, which prompted chemists to try tasting them (@mako-yass), and (2) maybe some of these origin stories are scientific folklore (@d0themath). Scientists, who are very concerned to get the description of physical reality right, are surprisingly cavalier about describing their own history in an accurate way.
I understand the softness of categories, and I don’t mind that you would use the available data to not put me in the Christian box. Some things that you don’t see are that I engage in Catholic practices, like going to mass (which is precisely why I canned an earlier draft and I’m writing again now).
If I gave the impression that Jesus is an iteration in general improvement of morality, then I mischaracterized my belief and my community’s: we believe that Jesus is God—whatever that means. I have to add the “whatever that means” because it seems like a doctrine that deliberately confounds logic, like the bit about Buddha here, when paired with Christians’ transcendent notion of God. If we thought of gods as giants who lived on Mt. Olympus, then one of them becoming physical like Zeus-the-swan wouldn’t be a problem, but we go out of our way to describe God as being more like Plato’s Zeus, which is everything that a limited, embodied, human being isn’t. Catholics emphasize saints as evidence of continuing improvement, and the apostles are often portrayed as not understanding what was happening, but Jesus (and Mary) are untouchable.
On the other hand, I look at stories like Matthew 15:27, in which a Canaanite woman appears to teach Jesus about tolerance—at the beginning of the story, it seems like he didn’t know. Most people I talk to say that it was like Socratic questioning—he really did know—but maybe the divine part of him is that he caught on and accepted the correction? While God-as-hypostatic perfection can’t learn and improve (being outside of time), God-as-a-human being can and this is what it looks like? That sort of consideration is in the “whatever that means” phrase I used above.
Okay, now on the point about not mentioning heaven: not many people that I know do. Whereas I had to clarify that we follow the Jesus-is-God doctrine—quite heavily, it’s a frequent topic—I usually only hear about heaven at funerals. While I’m sure that the people around me believe in it as “consciousness does not extinguish at death,” the subject of heaven and hell come with a heavy dose of “this terminology/imagery is metaphorical.” They’d be quick to point out that heaven (and hell) is not a “place” and I think some popes have made comments about hell being a state and not a place. (In particular, I remember one from the 90′s, but that would be a few popes ago.)
You’re right that the 2 Thessalonians letter sounds like demons eating your guts, and anything in the canonical set of books is considered as writing inspired by God (with or without their authors’ understanding)—it doesn’t matter that the writer claims to be Paul and might have not been Paul. (Attributing works to your group’s founder seems to have been more common in the ancient world. I think it’s not controversial that there were three “Isaiahs.”)
Two things about that, though: Catholics don’t put equal weight on everything in the Bible—they’re all above a certain threshold of importance, but not equally important—and there’s no actual fire and brimstone imagery in it, mostly just about being “shut out,” the kind of imagery that Jesus used, for instance, in the wise and foolish virgins parable (Matthew 25). Meanwhile, a lot of material that didn’t make it into the Bible but was influential in the early church did have more viscerally imagined rewards and punishments after death. It was an idea that was in the air at the time in Judaism (except for the Sadducees), and only got a little bit into the canonical Bible.
So, clearly, Catholics would hold that there’s some kind of good and bad afterlife, and most of what I’ve heard has that you’ll either be “with God” or “not,” and not being with God is the bad thing in itself, irrespective of any gut-eating demons. Depictions of heaven and hell like C.S. Lewis’s The Great Divorce are popular. (Napoleon Bonaparte is all alone in a huge mansion, repeating to himself that it was everyone else’s fault...)
As for “No one comes to the Father except through me,” I had never connected that to the afterlife before; it has always seemed more like a general coming-to-Jesus saying (in life, for the sake of living, not specifically the afterlife).
It is the case (my impression, which would be interesting to verify with a survey because it’s an easy-to-ask question) that Catholics believe that there are non-Christians “with God” after death, i.e. in heaven. Even if they have to weasel out of some suggestively worded biblical passages (e.g. “good people who don’t acknowledge Christ are mystically going to the Father through Jesus”), or not even try to explain it (e.g. “God will figure it out/above my pay grade”), there’s a strong cultural current against making God look mean. Or a trickster, as you said in your original question.
Personally—maybe you’ll consider me even less of a Christian because of this—I don’t see an afterlife as something that happens to us as individuals. When there is talk about what heaven is like—hypostatic union, non-glass darkly—it doesn’t seem like much psychological continuity with one’s living ego. I’m not quite the person who inhabited this body 20 years ago, since my mind has changed a lot and what defines a person apart from their mind? So if we become outside-of-time, in-union-with-God, experiencing reality in a totally different way, how is that even me? Even following standard doctrine, the vision of heaven seems to have been exalted to such a degree that it’s no longer relevant: what I recognize of what I am will die when my body dies, and maybe something mystical that I don’t recognize goes off and does something else. But these are my own private musings (which I might change or maybe feel I have a better understanding of later) and not representative of Catholics or Christians in general.
I had seen Tegmark’s four multiverses before, but relating it to this is something I hadn’t considered. The first level of multiverses, with different initial conditions but the same laws, is a lot like how I understand Pearl’s ensemble of possible worlds.
As for the universe being a mathematical structure, that seems to be pretty much what physicists assume by default. The formulations of string theory that I’ve seen (casually—I’ve never worked in that area) replace the space-time manifold with a non-geometric matrix, making it math all the way down.
Even if it is math all the way down, I would consider physical reality to be a different thing from mathematical reality because the physical fact is which mathematical structure is actual. The same way that Chalmers calls consciousness a “further fact” above and beyond material brains, the fact of which mathematical structure is the one that is our world, as opposed to just being logically possible, is a further fact above and beyond the mathematical possibilities.
But if that observer is in the universe, then there’s more in the universe than just the circle.
I was examining this universe from the outside. We can’t actually do that, though we act as though we do in the physical sciences. (One idea in the physical sciences that takes seriously the fact that experimenters are a part of the universe they observe is superdeterminism, and it’s one of the possible loopholes for Bell’s Inequality.)
Sorry that I didn’t notice your comment before. You took it the one extra step of getting kinetic and rotational energy in the same units. (I had been trying to compare potential and rotational energy and gave up when there were quantities that would have to be numerically evaluated.)
Yeah, I follow your algebra. The radius of the ball cancels and we only have to compare and . Indeed, a uniformly solid sphere (an assumption I made) rolling without sliding without change in potential energy (at the end of the ramp) has 29% rotational energy and 71% linear kinetic energy, independently of its radius and mass. That’s a cute theorem.
It also means that my “physics intuition trained on similar examples in the past” was wrong, because I was imagining a “negligible” that is much smaller than 29%. I was imagining something less than about 5% or so. So the neural network in my head is apparently not very well trained. (It’s been about 30 years since I did these sorts of problems as a physics major in college, if that can be an excuse.)
As for your second paragraph, it would matter for solving the article’s problem because if you used the ball’s initial height and assumed that all of the gravitational potential energy was converted into kinetic energy to do the second part of the problem, “how far, horizontally, will the ball fly (neglecting air resistance and such)?” you would overestimate that kinetic energy by almost a third, and how much you overestimate would depend on how much it slipped. Still, though, the floppy track would eat up a big chunk, too.
Sorry—I addressed one bout of undisciplined thinking (in physics) and then tacked on a whole lot more undisciplined thinking in a different subject (AI alignment, which I haven’t thought about nearly as much as people here have).
I could delete the last two paragraphs, but I want to think about it more and maybe bring it up in a place that’s dedicated to the subject.
It might not matter in the grand scheme of things, but my comment above has been on my mind for the last few days. I didn’t do a good job of demonstrating the thing I set out to argue for, that effect X is negligible and can be ignored. That’s the first step in any physics problem, since there are infinitely many effects that could be considered, but only enough time to compute a few of them in detail.
The first respondent made the mistake of using the challenger’s intentions as data—she knew it was a puzzle that was expected to be solvable in a reasonable amount of time, so she disregarded defects that would be too difficult to calculate. That can be a useful criterion in video games (“how well does the game explain itself?”), it can be exploited in academic tests, though it defeats the purpose to do so, and it’s useless in real-world problems. Nature doesn’t care how easy or hard a problem is.
I didn’t do a good job demonstrating that X is negligible compared to Y because I didn’t resolve enough variables to put them into the same units. If I had shown that X’ and Y’ are both in units of energy and X’ scales linearly with a parameter that is much larger than the equivalent in Y’, while everything else is order 1, that would have been a good demonstration.
If I were just trying to solve the problem and not prove it, I wouldn’t have bothered because I knew that X is negligible than Y without even a scaling argument. Why? The answer physicists give in this situation is “physics intuition,” which may sound like an evasion. But in other contexts, you find physicists talking about “training their intuition,” which is not something that birds or clairvoyants do with their instincts or intuitions. Physicists intentionally use the neural networks in their heads to get familiarity with how big certain quantities are relative to each other. When I thought about effects X and Y in the blacked-out comment above, I was using familiarity with the few-foot drop the track represented, the size and weight of a ball you can hold in your hand, etc. I was implicitly bringing prior experience into this problem, so it wasn’t really “getting it right on the first try.” It wasn’t the first try.
It might be that any problem has some overlap with previous problems—I’m not sure that a problem could be posed in an intelligible way if it were truly novel. This article was supposed to be a metaphor for getting AI to understand human values. Okay, we’ve never done that before. But AI systems have some incomplete overlap with how “System 1” intelligence works in human brains, some overlap with a behavioralist conditioned response, and some overlap with conventional curve-fitting (regression). Also, we somehow communicate values with other humans, defining the culture in which we live. We can tell how much they’re instinctive versus learned by how isolated cultures are similar or different.
I think this comment would get too long if I continue down this line of thought, but don’t we equalize our values by trying to please each other? We (humans) are a bit dog-like in our social interactions. More than trying to form a logically consistent ethic, we continually keep tabs on what other people think of us and try to stay “good” in their eyes, even if that means inconsistency. Maybe AI needs to be optimized on sentiment analysis, so when it starts trying to kill all the humans to end cancer, it notices that it’s making us unhappy, or whimpers in response to a firm “BAD DOG” and tap on the nose...
I’m still Catholic. I was answering your question and it got long, so I moved it to a post: Answer to a question: what do I think about God’s communication patterns?
I think it would be more correct to say that a focus on believing particular assertions is a fairly recent trend in religion, encompassing the past millennium or two, but really picking up in the last few centuries.
It happened in or between Christianity and Islam (as isusr points out), and they probably both influenced each other. For example, Protestant Christianity focuses a lot more on a holy book than Catholicism and Orthodox Christianity, but in a way that resembles Islam’s veneration of the Quran: citing verses to prove points. Since then, Catholics and Orthodox have also stepped up their focus on the Bible. There’s a lot of cross-pollination.
In the last century or so, religious statements have been presented as a kind of alternate-science (e.g. Young Earth science), presumably to respond to an apparent threat, but this is a very new way of taking about religion. There were biblical literalists (and non-literalists) throughout Christian history, but ancient theologians would probably accuse these people of missing the point.
Meanwhile, religions with only recent sustained contract with Christianity and Islam (past half-millennium) and religions that preceded them focus a lot more on practice, i.e. ritual and social behaviors. Some belief is implicit (e.g. why leave offerings for gods or ancestors if you don’t think they exist in a form that would benefit from the offerings?), but they are much less the focus.
You might not be able to wiggle it enough to get what you want (depending on what you want), but you need to be able to wiggle it, yes.
Given the way “A causes B” is used in everyday speech and among philosophers, it seems that it needs to have some notion of “if A, then B” (with a possible addition of “is more likely”). To use “casual” to mean “deterministic” would be confusing—different enough, anyway, that this usage needs to be called out as unrelated, to avoid confusion. “Smoking causes cancer” is not deterministic (some smokers are lucky) and includes the idea that not smoking is a possibility.
I know there’s the wildly different usage of “to cause” in relativity to mean “is before, with time-like separation,” and that is also terribly confusing. It should be called out more.
This is a good example of needing to watch my words: the same sentence, interpreted from the point of view of no-free-will, could mean the complex function of biochemical determinism playing out, resulting in what the human organism actually does.
What I meant was the utility function of consequentialism: for each possible goal , you have some preference of how good that is , and so what you’re trying to do is to maximize over . It’s presupposing that you have some ability to choose instead of , although there are some compatibilist views of free will and determinism that blur the line.
My point in that paragraph, though, is that you might have a perfectly rational machinery for optimizing , but one has to also choose . The way you choose can’t be by optimizing over . The reasons one has for choosing also can’t be directly derived from scientific observations about the physical world, because (paraphrasing David Hume), an “is” does not imply an “ought.” So the way we choose , whatever that is, requires some kind of argumentation or feeling that is not derivable from the scientific method or Bayes’ theorem.
And even more deeply than door-to-door conversations, political and religious beliefs spread through long-term friend and romantic relationships, even unintentionally.
I can attest to this first-hand because I converted from atheism to Catholicism (25 years ago) by the unintended example of my girlfriend-then-wife, and then I saw the pattern repeat as a volunteer in RCIA, an education program for people who have decided to become Catholic (during the months before confirmation), and pre-Cana, another program for couples who plan to be married in the church (also months-long). The pattern in which a romantic relationship among different-religion (including no-religion) couples eventually ends up with one or the other converting is extremely common. I’d say that maybe 90% of the people in RCIA had a Catholic significant other, and maybe 40% of the couples in pre-Cana were mixed couples that became both-Catholic. What this vantage point didn’t show me was the fraction in which the Catholic member of the couple converted away or maybe just got less involved and decided against being married Catholic (and therefore no pre-Cana). I assume that happens approximately as often. But it still shows that being friends or more than friends is an extremely strong motivator for changing one’s views, whichever direction it goes.
Since it happened to me personally, the key thing in my case was that I didn’t start with a clear idea of what Catholics (or some Catholics, anyway) actually believe. In reading this article and the ones linked from it, I came to Talking Snakes: A Cautionary Tale, which illustrates the point very well: scottalexander quoted Bill Maher as saying that Christians believe that sin was caused by a talking snake, and scottalexander himself got into a conversation with a Muslim in Cairo who thought he believed that monkeys turned into humans. Both are wild caricatures of what someone else believes, or at least a way of phrasing it that leads to the wrong mental image. In other words, miscommunication. What I found when I spent a lot of time with a Catholic—who wasn’t trying to convert me—was that what some Catholics (can’t attest for all of them) meant by the statements in their creed isn’t at all the ridiculous things that written creed could be made to sound like.
In general, that point of view is the one Yudkowsky dismissed in Outside the Laboratory, which is to say that physical and religious statements are in different reality-boxes, but he dismissed it out of hand. Maybe there are large groups of people who interpret religious statements the same way they interpret the front page of the newspaper, but it would take a long-term relationship, with continuous communication, to even find out if that is true, for a specific individual. They might say that they’re biblical literalists on the web or fill out surveys that way, but what someone means by their words can be very surprising. (Which is to say, philosophy is hard.) Incidentally, another group I was involved in, a Faith and Reason study group in which all of the members were grad students in the physical sciences, couldn’t even find anyone who believed in religious claims that countered physical facts. Our social networks didn’t include any.
Long-term, empathic communication trades the birds-eye view of surveys for narrow depth. Surely, the people I’ve come in contact with are not representative of the whole, but they’re not crazy, either.