Posts
Comments
Here are some examples, all said in a heartfelt, emotionally connected way: “I’m happy you are here”, “I start smiling when I see you, I like having you around”, “I like you”, “I like talking to you”.
I'm trying to imagine myself receiving this sort of compliments from someone other than a close friend, and I probably won't be very happy about it... "I start smiling when I see you" in particular is vaguely scary (while the others are just blunt and would leave me embarassed).
Yup, I know people among all age categories who basically never read books on their own. One of them was my grandpa (RIP), who once tried to read a long fiction book (I don't remember which one), managed to read one page a day with significant effort, and quitted shortly after.
We see similar patterns in the transition from Newtonian to relativistic mechanics or discursive Greek geometric algebra to symbolic Arabic equations or superstitious alchemy to physically grounded chemistry. There are thousands of other examples. It is not a general rule that all the past knowledge must be learned to create something new. Often, past knowledge is completely supplanted by a new discovery and progress can continue without increasing, and often decreasing, the necessary educational investment.
The Ptolemaic model is an extreme example. I doubt that we can actually find thousands of other huge corpus of accumulated knowledge who were later utterly trashed. Yes, students today do not learn the Ptolemaic model, because it was plain wrong. But the transition from Netwonian to relativistic mechanic is definitely not in the same reference class, since high school students today are still starting physics courses studying Netwonian mechanics, which are a very good approximation of relativistic mechanics in simplified conditions. You can't just dismiss Newton as superseded by Einstein (and I also doubt that Greek geometry is dead).
Moreover, if you sit in the frontier of knowledge developing a shiny new model, you can't just blatantly ignore the current model (even if it's wrong): that luxury will belong to future scientists. The developer of the new model must also master the old one in order to explain why the new one is better.
There's a thing in our very literate modern society which still survives following more or less the same pattern: jokes (and the longer ones in particular).
Think about it: even if you can easily find printed books full of jokes, in practice jokes are mostly an oral thing. You tell one to a friend, who in turn tells it to another friend, and so on. But unless the joke is a single sentence, at every step in the chain it will be distorted and retold a bit, even if it remains recognizably the same joke (for some people, adding a lot more words to the original version is also not uncommon). Nobody is expected to remember a joke word-by-word, and even the same person telling the same joke twice will probably not use the very same words. Yet it will be the same joke.
Maybe being a guslar is not so different from telling a joke 2294 lines long. Note also that "being able to repeat a joke non-exactly after hearing it once" is not considered difficult, even if the joke is 200+ words long (while "being able to repeat 200 words exactly" is basically impossible for the average person).
I find quite amusing for this post to have been published the same day as this smbc.
I'm sure that most cryptographers reading this post did not believe me when I mentioned that FHE schemes with perfect secrecy do exist.
This is a 2013 paper that was already cited 71 times according to Google Scholar... I'm not a cryptographer but I would bet that the average cryptographer is not fully clueless about this.
What does it mean to not even be trying?
It does not only mean the things Alexander pointed us to last time, like 62% of singles being on zero dating apps, and a majority of singles having gone on zero dates in the past year, and a large majority not actively looking for a relationship.
Maybe this does not apply to the Bay Area, but I find worth saying that if you live in the average town, being on a dating app comes with some social stigma attached, for both genders (much less than actively paying sex workers, I suppose, but still). I am on zero dating apps, but I'm pretty sure that my mother would scold me forever should I ever try that. Anyway, this has nothing to with the actual reason I'm not on dating apps, which is that I don't want a relationship in the first place (not in the usual "I stopped trying" sense, I literally mean "I never tried because I'm very happy being left alone"... can we please acknowledge that "no relationships" could be an actual preference for some people?)
If you are a paesant in USA/Europe from 200 years ago (or even 100 years ago), then you are very very likely to spend basically all your life in your home town, and your dating pool is restriced to a few dozens of people you know in person. Also is not uncommon for your parents to basically arrange your marriage themselves. The dating experience of my grand-grandmother (born 1899) was:
- your suitor talks to you a few times while you are walking back home from church
- your uncle closely follows you both to ensure nothing scandalous happen
- you are married shortly after (then you can start, you know, actually touching your husband)
Of course, the situation was different 50 years ago, but even then, your dating pool was mostly limited to the Dunbar-sized group of people you knew in person. Imagine to be in 1970 Wyoming. Maybe your perfect soulmate lives just a few miles apart in another town, but you have no reliable way to search them. And if your perfect soulmate lives in France (for some reason), you are not going to meet them full stop.
It's certainly a good imitation of average (i.e. bad) writing. I couldn't bear reading any of these stories past the first paragraph or two.
I agree, they feel very much like oversimplified stories aimed to 12-year old readers at best. Are "novice writers with some years of practice" actually worse than this?
You don’t only get C-3PO and Mario, you get everything associated with them. This is still very much a case of ‘you had to ask for it.’ No, you did not name the videogame Italian, but come on, it’s me. Like in the MidJourney cases, you know what you asked for, and you got it.
I consider this a sort of overfitting that would totally happen with real humans... I bet that pretty much anything in the training set that could be labeled "animated sponge" are SpongeBob pictures, and if I say "animated sponge" to a human, it would be very difficult not to think about SpongeBob.
I also bet that the second example had to use the word "droid" to do the trick, because a generic "robot" would have not been enough (I've never seen the word "droid" at all outside the Star Wars franchise).
I suggest another test: try something like "young human wizard" and count how many times it draws Harry Potter instead of some generic fantasy/D&D-esque wizard (I consider this a better test since Harry Potter is definitely not the only young wizard depicted out there).
Imagine you were choosing between two potential ~10min car journeys: one being 6mi and one being 200ft shorter but you're not allowed to use your phone, read a book, listen to music, etc. I think nearly everyone would chose the extra 200ft, no?
If you mean "as a passenger", then sure. Otherwise I would totally pick the first one (even without the 200ft discount), since you are very much not supposed to do these things while driving.
My anecdotal evidence from relatives with toddlers is that the first few years of having your first child is indeed the most stressful experience of your life. I barely even meet them anymore, because all their free time is eaten by childcare. Not sure about happiness, but people who openly admit to regretting having their kids face huge social stigma, and I doubt you could get honest answer on that question.
Is a photographer "not an artist" because the photos are actually created by the camera?
It could be a defensible position (surely held by many 19th century painters), but I suppose that many people would find significant differences between an actual professional photographer and a random bozo with a camera, even if you give both the same camera.
I have a similar feeling about AI artwork. As long as AI remains a tool and does not start to generate art on its own, there will be a difference between someone who spends a lot of time carefully crafting prompts and a random bozo who just types "draw me a masterpiece". I would consider fair to define as an "artist" the first one and not the second.
They aren't now, but if you look back at the time when sugar was actually discovered (around 1500), every European noble who could afford it did in fact revel in sugar. The wedding of Ercole d'Este, duke of Ferrara, is a particularly infamous example. I cannot find the historical menu in English, but the original Italian version is available here.
Something about this feels off to me. One of the salient possibilities in terms of technology affecting romantic relationships, I think, is hyperspecificity in preferences, which seems like it has a substantial social component to how it evolves. In the case of porn, with (broadly) human artists, the r34 space still takes a substantial delay and cost to translate a hyperspecific impulse into hyperspecific porn, including the cost of either having the skills and taking on the workload mentally (if the impulse-haver is also the artist) or exposing something unusual plus mundane coordination costs plus often commission costs or something (if the impulse-haver is asking a different artist).
It's even worse than this. Even if you restrict to super-mainstream porn, you can of course find a deluge of naked people doing naughty things, but it's very rare for these people to be the epitome of beauty. The intersection between "super duper hot" and "willing to appear in porn videos" is small, and nobody expects random camgirls to look like Jessica Rabbit (presumably because actual ultra-hot people have no difficulty finding any other job). Add just a simple preference for a specific ethnicity or the like, and Stable Diffusion rapidly becomes the only way to find photorealistic images.
"Humanity" is not a single agent. It can have preferences only insofar as the individual humans share similar preferences. If you are a happiness maximizer, for every individual human you look at, you'll probably find that their happiness would be maximized by wireheading (because most whimsical desires like becoming the king of the world are not feasible to satisfy otherwise).
I'm not even that sure that CEV would avoid this problem. In which way being enclosed in a perfect world is not the best thing for you? Because it would be fake? But how do you know that's fake? Imagine that an AGI offers to teleport you on another planet, which is perfectly suitable for you; you'll land there and thrive forever. Now imagine that instead of actually teleporting you to another planet, it just let you to believe the same; you'll (mentally) land there and thrive forever anyway. I mean, your brain is experiencing the very same thing! It's not obvious that the second option is worse than the first one, unless you have an hardcoded limit like "Thou Shalt Not Wirehead People".
Being embedded in a fake reality and fooled into believing it's true would be against many people's preferences.
Only if they can see through the illusion in the first place. Suppose that the happiness-maximizing AGI could throw you into a perfect fake reality without you even noticing. One day you wake up, and from that point on everything goes perfectly well for you in some very plausible manner. After a while, your happiness has reached immensely high levels. "This poor schmuck doesn't want to be wireheaded" is not a valid enough justification for not wireheading them in secret (for comparison, consider an angry teenager who break up with their partner and firmly says they want to immediately die.. you are probably going to ignore the stated preference and preventing them from suicide, knowing that this will result in much higher future happiness).
Is an utopia that'd be perfect for everyone possible?
The short and obvious answer is no. Our civilization contains omnicidal maniacs and true sadists, whose central preferences are directly at odds with the preferences of most other people. Their happiness is diametrically opposed to other people's.
If you are constrained to keep omnicidal maniacs in the same world as everyone else, this is obviously true.
But it doesn't seem to be obviously true in every possible future. Imagine a world where every single sentient mind is wireheaded into their own personal paradise, interacting only with figments designed to maximize their preferences.
Well, for what's worth, I can write a symphony (following the traditional tonal rules), as this is actually mandated in order to pass some advanced composition classes. I think that letting the AI write a symphony without supervision and then make some composition professor evaluate it could actually be a very good test, because there's no way a stochastic parrot could follow all the traditional rules correctly for more than a few seconds (an even better test would be to ask it to write a fugue on a given subject, whose rules are even more precise).
The video was recorded in 2016, 10 years after his 2006 injury. It's showing the result of 10 years of practice.
Ok, fair enough. But he started playing in concerts long before 2016, and the first recorded album was released on February 2007. Apparently he was selected as the 2007 Independent Artist of the Year by the LA Association of Independent Artists, which seems still quite impressive for someone starting to play less than one year before (is this a real association? I've never heard of it before).
Yeah, other people already pointed this out. I blame writers who cannot do math here. I know people in the humanities that would happily define a "math genius" anyone capable of correctly calculating an integral.
Anyway, even taking for granted that he's just a crank who draws some cool triangles, it's still quite impressive that an head injury could turn a normal guy into a math crank.
Ok, maybe I shouldn't have used the same words used by clickbait youtube videos.
Anyway, he seems more interested in drawing triangles than studying math textbooks, so I don't expect him to produce novel insigths. On the other hand, plenty of people are very good at math but never produce any technical writing on scientific journals. If banging your head can bring you from 50° percentile to 90° percentile in math attitude, that's still pretty big news even if you don't literally become a math genius (his story seems to strongly imply that his past self wouldn't have been able to pass those math classes).
You are very welcome to investigate! I don't deny that all of this is very perplexing. But it is at least plausible (in the sense of "not requiring to break laws of physics") than a head injury could have a one-in-a-million chance of modifying your brain in strange ways.
Consider that this guy holds public concerts since 2007, we have no way to deny that he can actually play piano (albeit in a strange, untrained way). The only way for this to be a scam would be for him to have trained alone in complete secret for something like 10 years.
Also, he claims to "see" music in a way that reminds me of one historical anecdote about Mozart (he also claimed to "see" whole symphonies compressed into points, waiting only to be unraveled).
Apparently someone did exactly this in 2009. Could someone more familiar than me with the relevant literature have a look?
Have you ever played piano?
The kind of fluency that we see in the video is something that a normal person cannot acquire in just a few days, period. Even if he didn't literally play perfectly the first time, playing perfectly after one month would still be incredibly impressive. You plain don't become a pianist in one month, especially without a teacher, even if you spend all the time on the piano.
Also, this guy is apparently still not able to read sheet music and still doesn't know anything about music theory. It's difficult to explain in non-technical terms, but his music is exactly the kind of music that I would expect from an incredibly talented person who knows nothing about music theory.
I think the thing you're missing is you're still exposed to crashes because of some maniac doing something extremely risky and hitting you.
Yes, but this is true even when I'm not driving. An out-of-control car could crash into me even when I'm walking or sitting inside a bus (and in some cases even when I'm at home).
Anyway, thanks, I'll look into this paper.
This is the same chart linked in the main post.
Again, I am not here to dispute that car-related deaths are an order of magnitude more frequent than bus-related deaths. But the aggregated data includes every sort of dumb drivers doing very risky things (like those taxi drivers not even wearing a seat belt).
Since I'm quite confident to have a particularly cautious driving style, I am not very interested in the total number of fatal car accidents, because lots of people driving recklessly make it skyrocket (I'm not relying entirely on self-judgement here; anecdotally, at least four times I gave a ride to a friend and they mocked me for my overly cautious driving).
To make a comparison, take this document on cancer incidence (chart on page 2). Lung cancer is the most frequent cancer of them all, so you should be more worried about lung cancer than every other cancer, right? Wrong, unless you smoke, since people who have never smoked only make up 10-20% of all lung cancer cases (it follows that you should be 5x less worried about lung cancer if you aren't a smoker, I presume).
I'm now trying to find data on the number of car fatalities not involving people doing stupid things like texting or speeding. I thought that taxi drivers were a good proxy for cautious drivers, but I was very wrong and now I don't know what else to use as a proxy.
Table 10 also shows that some 30% of taxi drivers involved in crashes weren't wearing seat belts (they're apparently not legally required to in NSW! news to me), which is a pretty big clue that taxi drivers aren't the paragon of careful driving one might assume.
WTF!?
Ok, I suppose I have to update my priors on taxi drivers (man, they even write "There is considerable anecdotal evidence that taxi drivers around the world drive in a manner the rest of the public considers to be unsafe").
Do you have suggestions about other proxies for careful driving?
"Cautious driver" is not a real category. It's not something my crash database can filter on.
Yes, obviously it is not a well-defined category, I mostly hoped that you could filter for taxi or similar.
Anyway, I am not claiming to be the best driver in the world (although I'm 100% safe at least w.r.t. drinking since I don't drink at all), I'm just claiming to be at least as good as a taxi driver, and I would be really really surprised if it turned out that taxi drivers crash their vehicles with the same frequency as the general population.
The thing about human error is that you make errors ALL THE TIME. You, or other road users, should not die because of your errors. And the errors that tend to result in fatal crashes are not "I was drunk and on meth and speeding" (though those obviously do), the ones that more commonly result in fatal crashes are "I looked away from the road for a second to adjust my GPS and hit a pedestrian".
Well, averting your eyes from the road and your hands from the wheel at the same time in order to touch the screen (rather than reaching a calm spot and stop the car first) is so obviously risky that I was lumping it into "idiots looking at their smartphones".
Yes, I do know that even a minor distraction could be enough to crash your car. My point was that also bus drivers are humans and make errors all the time. I generally drive alone, in silence and with the phone turned off, in order to maximize my attention, and it doesn't seem obvious that a random bus driver should have a lower probability to crash their vehicle than me (given that the average bus driver is also older than me and drives a huge vehicle which is surely more difficult to handle than my small car).
Do you have actual numbers on fatalities caused by cautious drivers?
My guess is that most people's intuitive sense of the danger of cars versus trains, planes, and buses has been distorted by this coverage, where most people, say, do not expect buses to be >16x safer than cars.
Well, even after eyeballing this graph, I still don't expect to be 16x safer on a bus than while driving my car.
My experience is that car crashes are covered at least by local news, and the overwhelming majority of car crashes I've heard of involved drunk drivers, ludicrous speed, or idiots looking at their smartphones instead of the road. A bus is safer than a car mainly because the average bus driver is more scrupulous than the average car driver.
Do we have data on car crash fatalities limited to public services like taxi? My best guess is that fatalities would decrease by an order of magnitude when you restrict to rule-abiding drivers.
Uh, this is somewhat surprising.
Do you mean that after your personal growth, your social circle expanded and you started to regularly meet trans people? I've no problem believing that, but I would be really really surprised to hear that no, lots of your longterm friends were actually trans all along and you failed to notice for years.
As I said in other comments, I am not locked in some strange conservative bubble keeping queer people out. For instance, I know at least three lesbians: one of them is a very obvious butch lesbian always dressed in male clothes, the other two are not so obvious but I still guessed they were lesbians quite early (say, around the third or fourth encounter in both cases). And I am surprised because this never happened with trans people, in the sense that I never caught the slightest hint that one of my longterm acquaintances could possibly be born with a different gender.
The problem with this analogy is that megacorps must at least pay lip service to the rule of law, and there's no way a megacorp would survive if the government decide that they shouldn't. Any company is ultimately made of people, and those people can be individually targeted by the legal system (or worse). What's the equivalent for AGI?
Ok, but I still think it's legit to expect some kind of baseline skill level from the human. Doing the deceptive chess experiment with a total noob who doesn't even know chess rules is kinda like assigning a difficult programming task with an AI advisor to someone who never wrote a line of code before (say, my grandma). Regardless of the AI advisor quality, there's no way the task of aligning AGI will end up assigned to my grandma.
I've mixed feelings about this. I can concede you this point about short time control, but I am not convinced about notation and basic chess rules. Chess is a game where, in every board state, almost all legal moves are terrible and you have to pick one of the few that aren't. I am quite sure that a noob player consistently messing up with notation would lose even if all advisors were trustworthy.
It's probably better to taboo "talking" here.
In the broader sense of transmitting information via spoken words, of course that GPT4 hooked to a text-to-speech software can "talk". It can talk in the same way Stephen Hawking (RIP) could talk, by passing written text to a mindless automaton reader.
I used "talking" in the sense of being able to join a conversation and exchange information not through literal text only. I am not very good at picking up tone myself, but I suppose that even people on the autism spectrum would notice a difference between someone yelling at them and someone speaking soberly, even if the spoken words are the same. And that's definitely a skill that GPT-conversator should have if people want to use it as a personal therapist or the like (I am not saying that using GPT as a personal therapist would be a good idea anyway).
It's probably worth mentioning that 2/3 of all Wikipedia contribution are done by its 1000 most active users, with single persons like this guy doing millions of edit. I suspect that people with OCD are a supermajority among Wikipedia greater contributors.
I didn't try it, but unless your GPT conversator is able to produce significantly different outputs when listening the same words in a different tone, I think it would be fair to classify it as not really talking.
For example, can GPT infer that you are really sad because you are speaking in a weeping broken voice?
Most trans folk are still not transitioning.
But man, it only takes one.
Are the base rates for actual transition so low that you can have thousands of people in your extended social circle and still never hear of one?
I mean, being told by anyone "Hey, do you know that guy? Is a woman now." would be enough. But it never happens! And this is the kind of rare gossip that I would expect most of my acquaintances to share, especially the least trans-friendly ones (like my grandma, who's a devout Catholic and knows half the town by name).
Of course that trans people will go to the same grocery store and gas station as me! But for some reason I have zero of them in my social circle, like Scott has zero creationists (and the base rate for creationists is way higher than the base rate for trans people). Are we implying that no, at least some of Scott's friends must be closeted creationist?
Just one example: are you familiar with the concept of people starting a conversation by specifying the pronoun they want used to refer to them?
I'm not. In the sense that I've literally never ever met one of these people. I read about them on the Internet every time I open rationalist-related media, but they may as well be made of dark matter. I never meet them in everyday life!
I am interacting with the Rationalist community basically through LessWrong only. As you could infer from my first post, I don't live near SF or other notoriously queer-friendly cities, and I've never been to a Rationalist meetup in person. This is kind of the point: the average joe in the average mid-sized town could live for years and years without meeting a single obviously trans person (in the sense of someone who had visibly transitioned through medical procedures).
Right, I don't claim to be able to spot trans people who didn't start the transition, but at least for those who finished the transition, I assume that a prolonged interaction would at least reveal some clues. Take, I don't know, my conservatory (at least 60-80 people I personally interacted with for years, including some of those gays and lesbians from the previous post). Even if with these people I talk mostly about music, I would be truly shocked to find out that one of them was trans all along.
Do you want larger numbers? My father runs a small business with ~1000 customers, and most of them have been the same for years. Even if he doesn't personally know all of them, I am quite sure that he would notice if one of them transitioned. So far, he has not.
I can concede that maybe I've walked near trans passers-by who didn't obviously look like trans people, but I'm still confident that 100% of people I interacted with verbally more than once are not trans. I suppose that homosexuals could pass as straight more easily than trans could pass as cis, but I did meet gays and lesbians nonetheless (indeed, most of them don't obviously look like homosexuals).
To the best of my knowledge, there are no LGBT organizations in my town, but there are certainly some in the bigger city where my workplace is located. I've no doubt I would find a trans person there if I went looking. My point was that, despite having interacted with hundreds of people at this point, I've never met one by chance in the same way I met gays and lesbians.
I like this post, but I think that it underestimates a bit the strength of filter bubbles. Even if a trait is present in 1% of the general population, it doesn't automatically follow that literally everyone has someone in their social circle with that trait.
If you walk around aggressively signaling the equivalent of “What!?? There ain’t no faggots in this town,” then it should be no surprise to you that your lived experience is that none of the people around you are gay (even though base rates imply that quite a few of them are).
I consider myself liberal enough to not project an anti-trans aura, but I still never meet trans people in person, like Scott Alexander never meet conservatives. Base rates seem to imply that there should be dozens of trans people in my town, but I've never seen one, and I don't know of anyone who has. My best guess is that when they decide to change sex, they also relocate to a more queer-friendly place. Taking SF as an extreme case, in the Bay you can find companies entirely run by trans women, a thing that I'm pretty sure never happens in other towns.
Hu, I don't know. I can totally imagine being mildly annoyed by this, especially if they end up inviting me to a social event on short notice.
Hu, actually I never tried just the face, I needed at least the upper torso and preferably the full figure.
Anyway, I spent a few hours today toying with that generator (I previously used mostly this). A very simple prompt like "An elf made out of green metal" can produce a somewhat okay result, but the elf will be either naked or dressed head to toe in green. You can try to add more bits to the prompt in a controlled manner: hair color/hairstyle, outfit/dress color, and the like, but the more details you add, the more the model is prone to forget some of them, and the first to be forgotten is often the most important (being made of green metal).
To be clear, the success rate is not 0%. I was eventually able to obtain an image kinda resembling what I wanted, but I had to sit through >200 bad images and it definitely wasn't an easy task. For these kind of things, we are totally not at the point where image generation "just works" (if you instead need a generic fantasy elf, sure, then it just works on the first try).
I will add that even taking humans aside, the remaining comparisons seem still quite bonkers to me. 1 carp ~ 1 bee sounds really strange.
Indeed, but an image generator is supposed to be useful for something other than generating an endless scroll of generic awesome pictures with wonky details; this kind of thing becomes boring really quickly. What most people actually need from an image generator is a sufficiently good replacement for the drawing skill they don't have.
To be clear, I share Portia's frustrarion here. I've been trying to get image generators to generate DnD portraits for months, and if the character is something more complicated than a Generic Tolkienian Elf or similar, you have to play increasingly complex shenanigans to obtain passable results. For example, I really really couldn't convince the AI to generate an elf literally made of green metal rather than just dressed in green (this was supposed to represent the effect of a particular prestige class turning the character into a construct).
I have a similar feeling... receiving a full-face smile out of the blue from a total stranger is vaguely creepy