Posts
Comments
This is more regimented than anyone I know, except for a few immune-compromised friends (and they aren't relaxing much based on current trends).
FWIW I've been a bit stricter, and I assume a chunk of LessWrongers have been as well.
I agree. I'm not sure how much to update on the things you mention or on other things that have happened since 2013, so I think my answer serves as more of jumping off point than something authoritative. I edited it to mention that.
See FAQ #4 on MIRI's website below.
Edit: It was written in 2013 so it is probably best viewed as a jumping off point from which you can make further updates based on what has happened in the world since then.
In early 2013, Bostrom and Müller surveyed the one hundred top-cited living authors in AI, as ranked by Microsoft Academic Search. Conditional on “no global catastrophe halt[ing] progress,” the twenty-nine experts who responded assigned a median 10% probability to our developing a machine “that can carry out most human professions at least as well as a typical human” by the year 2023, a 50% probability by 2048, and a 90% probability by 2080.5
Most researchers at MIRI approximately agree with the 10% and 50% dates, but think that AI could arrive significantly later than 2080. This is in line with Bostrom’s analysis in Superintelligence:
My own view is that the median numbers reported in the expert survey do not have enough probability mass on later arrival dates. A 10% probability of HLMI [human-level machine intelligence] not having been developed by 2075 or even 2100 (after conditionalizing on “human scientific activity continuing without major negative disruption”) seems too low.
Historically, AI researchers have not had a strong record of being able to predict the rate of advances in their own field or the shape that such advances would take. On the one hand, some tasks, like chess playing, turned out to be achievable by means of surprisingly simple programs; and naysayers who claimed that machines would “never” be able to do this or that have repeatedly been proven wrong. On the other hand, the more typical errors among practitioners have been to underestimate the difficulties of getting a system to perform robustly on real-world tasks, and to overestimate the advantages of their own particular pet project or technique.
Given experts’ (and non-experts’) poor track record at predicting progress in AI, we are relatively agnostic about when full AI will be invented. It could come sooner than expected, or later than expected.
Experts also reported a 10% median confidence that superintelligence would be developed within 2 years of human equivalence, and a 75% confidence that superintelligence would be developed within 30 years of human equivalence. Here MIRI researchers’ views differ significantly from AI experts’ median view; we expect AI systems to surpass humans relatively quickly once they near human equivalence.
Sure thing.
Sorry, misread your comment. I just figured out the issue I'm having with the Add Tag button. It's that the middle of it is clickable but the surrounding area is not. Ie the <a> tag is clickable but the surrounding <span> is not.
I'm on a 2017 Macbook Pro using macOS 10.15.5. Initially I was on Firefox. On Chrome and Safari I get a "Error: TypeError: Cannot read property 'viewUnreviewedComments' of null" error instead. In all three cases the error appears in red text on the screen in place of "Show full sidebar".
Comments:
- I like the new light-grey background. It looks nicer aesthetically to me. I didn't notice any eyestrain before but now that you mention it I do notice some with the solid white background.
- This isn't new, but it's been a while since I've been logged out. Looking at the screenshots in this post, I've been reminded that the logged out page has some images at the top. It strikes me as noticeably more aesthetically pleasing when there are some pictures at the top. I don't mean to imply that there should be pictures at the top, just that seeing them looks a little nicer to me.
- I really like the tooltip when you hover over "Recommendations". I guess it's no big deal, but it feels cool to know what it's actually doing.
- I actually never noticed "Show tag filters" on the current site. I guess I'm just so used to coming to LessWrong that there's a sort of "muscle memory" where I do my thing and don't notice things that aren't part of my routine, if that makes sense. Now that I notice the tags they seem cool and I'm excited to use them. I really like that it could be weighted instead of binary. I feel like I probably wouldn't use them if they were binary because, for example, I don't want to ignore all AI posts, so hiding those posts would be too strong an action for me.
- This sounds small as I type it out but actually is pretty annoying. When I scroll down, there's a ton of tooltips that pop up and get in the way. I guess the only difference are the tooltips for the tags, but maybe that's an important difference. 1) The tooltip when you hover over a post isn't too bothersome to me because it's located far enough to the right where I can still read other post titles, so it isn't really getting in the way. But the ones for the tags do get in the way a lot. 2) The current tooltips are still a little overwhelming though, so maybe adding more just kinda puts things over the edge in a sense. (I know I said I liked the one for the "Recommendations", which might be contradictory, but I'm trying to give more of a stream of consciousness type of feedback.)
Bugs:
- When I click "Show full sidebar" I get a "Error: TypeError: p is nul" error.
- Tooltips overlap each other sometimes.
- Clicking the "+" to add a tag filter doesn't work.
This post makes me think about Non-Expert Explanation. The point being made in Non-Expert Explanation is that sometimes a random person's explanation of something will be what makes it click for you, and so it's useful to have different people explain a given thing, even if some of those people aren't experts. I agree with that idea and I've found it useful when I'm learning something to start off skimming through a bunch of different resources: blog posts, forum discussions, YouTube videos, etc., and then dive more deeply for the ones that clicked for me.
Ah, I see what you mean. Although my understanding of the EA movement is that it's very acceptable, or even advisable to spend enough on yourself so that you're reasonably comfortable, ie. living a Mustachian lifestyle. In which case the difference is just a matter of timing: a Mustachian first puts their money towards their own retirement and then puts it towards EA, whereas an EA puts it towards both in parallel. That doesn't seem like a large difference in philosophy to me. What do you think? Am I thinking about this the wrong way?
after you're a millionaire yourself (or however you define FI)
This is tangential, but $1M seems like it's on the high end. I think that a lot of Mustachians would be retiring at more like $500-700k.
I think the piece does point to some interventions that we could deploy right now that would improve high-level decision making. For example, Experiment With Promising Ideas:
My home civilization, as you might guess, makes a huge deal out of the virtue of being fast to adapt: fast to respond to facts, fast to update and change policies. And if Sinyelt was right, the cost of people moving geographically was interfering with that virtue. So they built a test city - dath ilan had a concept of “let’s build a test city” - where houses were mated to modular foundations.
The line between what is the fruit of a saner society, and what is an intervention that will lead to a saner society seems blurry to me though. You could argue that Experiment With Promising Ideas is something that is impractical because we're not sane enough yet, and we have to get more sane first before trying to implement it. Or you could argue that it's something that we are capable of doing right now, and that it's part of the path towards sanity.
Here are some other excerpts from My April Fools Day Confession that might address your question:
- It’s not the evidence-based massage therapists who’ve been iterating their art with randomized experiments and competitions for 350 years
- In the world of dath ilan, everyone learns at age 9 about Nash equilibria, and there is a concept of a making a collective and virtuous effort to get past them. So as soon as computers and batteries were good enough to autopilot electric cars in a system of tunnels, the thing was done.
- And now I’m talking about how the economy worked, so I’ll go ahead and talk about some other things that dath ilan considered obvious. The medical profession was divided into junior diagnosticians, whose main job was to diagnose the obvious and know when the obvious had been called into doubt; and senior diagnosticians, who were highly paid and high-IQ and shadarak-trained, who could apply Bayes’s Rule in their sleep, and memorized all the prior probabilities, and had computers, and were graded on their probability calibrations.
- By which I mean that there would be centralized development of movies you watched on your own, and the training-games you played in what I won’t insult by calling it a school, and experiments to find out which variations worked.
- And even with respect to thorium power plants, China could offer a billion dollars in prize money to whichever group submitted a workable design for a liquid flouride thorium reactor, and another billion to the best iteration of that design, and another billion dollars divided up among people who found bugs, and then Earth’s civilization would have effective nuclear power plants. It was a bit more complicated than that in dath ilan, though it’s not like I was ever wise enough to look up the details. But I know that dath ilan had an established system for dividing prize money among contributors and checking for errors.
Eliezer had a lot of interesting ideas in My April Fools Day Confession, where he talked about a fictional society called Dath Ilan.
Looks like no one has brought up yet the tension between Mustachianism and EA... and then retires to do less or no work for pay, because their priority is freedom.
MMM only advocates less or no work for pay after retirement to the extent that work for pay makes you unhappy. See First Retire... Then Get Rich. He also is a fan of EA. More specifically, he supports the idea of first saving up your money so that you can retire, and then spending your time on EA once you're retired.
The Mustachian riposte to this is that, even if you want to work, isn't it better if you don't have to?
Right. MMM talks about this in Retire in your mind even if you love your job.
I think the honest description of their movement is that it's not about retiring early, which is only a true goal for very few of us--it's just about getting rich.
I think his position is that it's ultimately about the freedom to do what makes you happy. Most of the time your normal career isn't that, but if it is it makes sense to just retire in your mind. For everyone else, it makes sense to keep busy doing what you enjoy, whether or not that involves making money.
I think you may be misunderstanding his claims here.
He lost me a bit when he argued, quite vehemently, against having a pet.
In the Great News! Dog Ownership is Optional! article he doesn't argue against having a pet, just that "the benefits are generally understood, while the downsides and costs are vastly underestimated", and thus a lot of people may want to reconsider it.
He really believes that stripping every ounce of joy, except for what HE considers joyful, is the path to success.
I think this is easy to misinterpret because he doesn't do a good enough job of emphasizing this, but he doesn't believe in stripping joy away from life. I think this article on housing is a good example of that. He talks about why he thinks his nice four bedroom, four bathroom house is worth the large expense.
Yeah, that makes a lot of sense.
I get the impression that MMM does agree with paying for certain expensive things. For example, friends and family:
For slightly older and more settled folks like the MMM family, moving is temporarily off of the menu
and a home:
Or you could buy a used RV for $20k and cut your housing costs down by 90% for life.
But I wouldn’t do it, so I can’t tell you to do it. The Mr. Money Mustache way is not about living on the cheap. It’s about living the GOOD LIFE on the cheap. The fundamental lesson of this blog is that there is plenty of money to go around in this country, so you don’t have to eliminate your spending on everything to become financially independent. You just have to cut out your waste. And for the most part, buying yourself a home is not a waste.
I live in one of the nicer houses in my town’s nicest (to me) neighborhood
However, I agree that he doesn't do a good job of emphasizing these sorts of points, and I think that it would be good if he did emphasize them more.
Yeah, I think this post has helped me to understand the issues with blind empiricism. Thinking about it in terms of high-vs-low dimensional is really helpful to me.
Good for you!
If you haven't heard of him, I'd recommend checking out Cal Newport's writing, particularly Deep Work and Digital Minimalism.
This post seems like a good illustration of the value of Trying Things, particularly when the upside is high and the risk is low. It may seem obvious, but I know that I personally have a lot of low hanging fruit in this area that I haven't acted on for whatever reason.
So there's potentially a difference between abstractions which throw away basically all the structure vs abstractions which retain some.
Yeah, that's what I'm getting at.
Yeah I think your descriptions match what I was getting at.
Ah, I think I see what you mean. That makes sense that the high level model of the street map is also a summary statistic, not just the low level model with stuff thrown away. Let my try to refine my comment.
For the ideal gas example, I think of the low level model as looking something like this:
class LowLevelGas {
Particle[] particles;
}
class Particle {
String compound;
int speed;
int direction;
int mass;
// whatever else
}
And I think of the high level model as looking like this:
class HighLevelGas {
int pressure;
int volume;
int temperature;
}
LowLevelGas and HighLevelGas just look like there's a big difference between the two. On the other hand, LowLevelStreetMap and HighLevelStreetMap wouldn't look as different. It'd be analogous to a sketch vs a photograph, where the difference is sort of a matter of resolution. But with LowLevelGas and HighLevelGas, it seems like they are different in a more fundamental way. They have different properties, not the same properties at different resolutions.
I wonder if this "resolution" idea can be made more formal. Something along the lines of looking at the high level variables and low level variables and seeing how... similar?... they are.
Elizer's idea of Thingspace comes to mind. In theory maybe you could look at how close they are in Thingspace, but in practice that seems really difficult.
Hm, it seems to me that there is a distinction between 1) hiding information (or encapsulating it, or making it private), 2) ignoring it, and 3) getting rid of it all together.
-
For
setPasswordperhaps a programmer who uses this method can't see the internals of what is actually happening (the salting, hashing and storing). They just calluser.setPassword(form.password)and it does what they need it to do. -
For
User, in the example you give withList<User>, maybe we want to count how many users there are, and in doing so we don't care about what properties users have. It could beemailandpassword, or it could beusernameanddob, in the context of counting how many users there are you don't care. However, the inner details aren't actually hidden, you're just choosing to ignore it. -
For ideal gasses, we're getting rid of the information about particles. It's not that it's hidden/private/encapsulated, it's just not even there after we replace it with the summary statistics.
What do you think? Am I misunderstanding something?
And in the case that I am correct about the distinction, I wonder if it's something worth pointing out.
I wonder whether it'd be useful to distinguish between the following things.
Consider the example of the street map. is the exact same thing as except that there is detail removed (in a loose sense at least; in practice the map will probably have small differences compared to the territory).
Now consider the example of an ideal gas. throws away all of the stuff in and replaces it with summary statistics.
Both fit the definition of abstraction because you're removing information but maintaining the ability to answer questions, but in the ideal gas example you're adding something new in. Namely the summary statistics.
(Well, maybe "new" isn't the right word. Maybe it'd be better to say "adding summary statistics". I guess the thing I'm really trying to point at is the fact that something is being added.)
(I have a sense that the answer to this question is in the post but I'm having trouble extracting it out.)
There's something that I think of as composition, and I'm not sure if this fits the definition of abstraction. Consider in the context of programming a User that has email and password properties. We think of User as an abstraction. It's a thing that is composed of an email and password. I'm not seeing how this fits the definition of abstraction though. In particular, what information is being thrown away? What is the low level model, and what is the high level model?
The User example demonstrates composition of properties, but you could also have composition of instructions. For example, setPassword might consist of 1) saltPassword, 2) hashPassword and then 3) savePassword. Here we'd say that setPassword is an abstraction, but what is the information that is being thrown away?
(Typo: First bullet point under "Ideal Gas" should use instead of .)
Elizabeth's Epistemic Spot Checks series comes to mind.
SelfControl is also good for this.
Buying my girlfriend a vape. She smokes weed a lot and it's been said that getting a smoker to switch to vaping is one of the best things that you can do in terms of longevity. She also says that it is much more enjoyable.
(She likes the Pax 3 a lot; Vape Critic seems like a good resource for reviews.)
Is it possible that there are different viruses we're dealing with here? It just doesn't make sense to me that we're seeing such varying death rates, eg. Italy vs South Korea. The difference in medical quality doesn't seem like it can explain it.
It's been a while since I read REMOTE, sorry.
I haven't been able to personally try or validate much from the book. It's more that the things in the book make a lot of sense to me, and that I have a good amount of trust in the authors.
That said, there are some things that I do have personal experience with and can contribute my data point. I just started an actual remote job three weeks ago, and before that I've spent years as a solo founder of a startup, and autodidacting.
- The biggest thing (by far?) I've encountered is that it's important to have an off switch. Working from home, it can be tempting to check in and do a few tasks at 10pm. But when I do that, it makes it hard for my mind to properly "shut off" and relax.
- I've found cabin fever to be a minor issue when I stay home too much, but never a major one.
- I don't feel like I can get away with slacking off at home. I feel like I'm ultimately just being judged on my output, the same as it is in a physical job. At the end of the day if my output isn't there, I feel like my job would be at risk.
- At the job I'm currently at we do something called donuts where every two weeks we're paired with someone on the team to have a video chat with to get to know each other, and where talking about work is off-limits. The donut call I had made me feel closer to the people on the call, but I've also started to feel closer with the people I've been interacting with in general. And it's not clear to me that feeling closer to people translates to more productivity at all.
- Some people on my team aren't native english speakers and have trouble with writing, but despite that, I don't think it actually is much of a barrier. It involves a little bit more back and forth, but soon enough the signal comes through. This goes against one of the chapters in Remote. Perhaps being a clear thinker is what is important versus being a clear writer.
I'm a fan of REMOTE: Office Not Required.
Yes, exactly. Thank you for clarifying. I just read my original comment again and I think I didn't make it very clear.
What? Why? How do you decide which professionals to trust?
I was telling my friends and family to prep for the coronavirus very early on. At the time the main response was, "ok, chill, don't panic, we'll see what happens". Now that things have gotten crazy they think it's impressive that I saw this coming ahead of time. That's what my thinking was for point #3: perhaps this sort of response is common. At least amongst some non-trivial percentage of the population.
If you think your audience just isn't smart enough to evaluate arguments, then, gee, I don't know, maybe using a moment of particular receptiveness to plant a seed to get them to open their wallets to the right professionals later is the best you can do? That's a scary possibility; I would feel much safer about a fate of a world that knew how to systematically teach methods of thinking that get the right answer, rather than having to gamble on the people who know how to think about objective risks also being able to win a marketing war.
I very much agree, but it seems overwhelmingly likely that we live in a world where we can't rely on people to evaluate the arguments. And we have to act based on the world that we do live in, even if that world is a sad and frustrating one.
First, no one is choosing the virus so not a great comparison.
Yeah, that's true. When someone eats fast food every day and dies of a heart attack it's not quite as sad as when someone more innocent gets hit by a car.
But I view this as more of a mental skill that is built-up rather than something that people start doing immediately when thrust into lower-standad-of-living situations.
That's a great point. I got caught up thinking about how (I think) people should respond as opposed to thinking about how it'll actually play out in practice. That moves me a few more steps towards thinking that it is more harmful.
The fourth argument is just relevance to all of our wellbeing.
My intuition is that from here on out it's going to be hard to find steps we can take that will have even a moderate impact on our wellbeing.
1) We know that we need to avoid contact with others, so I assume we'll all being staying home. Given that we're at home isolated from others, is there much left to do? Things that go beyond common sense and standard advice, like opening packages outside and disinfecting them?
2) Eventually we'll face the question of when it is safe to end the quarantine. A conservative answer to that question is probably going to be "a few months after everyone else does". Maybe by studying it we'll learn that it's safe to end quarantine after two months instead of three, but that doesn't seem like it's a particularly impactful use of time.
3) Sadly, we can probably expect some members of our community to be infected. Or at least the loved ones of some members of our community. So then, the question of how to deal with infection is inevitably going to present itself.
I feel torn about whether that will be the most important thing to focus on when it does. On the one hand, when you shut up and multiply, I'm pretty sure that xrisk is many, many times more important. On the other, I really care about people in this community. I've always felt torn about this question of how much extra moral weight to give to those who I care about.
Regardless, I feel pessimistic that there will be much room for us to offer useful advice here. The big question is probably going to be whether you'll be able to navigate through the swarms in the hospitals to get access to treatment, and it seems unlikely that we'll be able assist with that.
Fortunately our community tends to be on the young side, and we are probably all quarantined by now, so we'll at least be good in a relative sense.
I think this made LessWrong a natural Schelling point of attention
Outsiders are paying attention to our coverage of the coronavirus? To a significant degree?
Do you know of places that would make good use of donations? If so, I strongly encourage you to write them up, ideally as answer here.
No, I don't know of places that would be particularly good to donate to. The WHO seems like a safe bet. Also, GiveWell is looking in to it (which I also noted in the other post).
Personally I suspect that even without knowing the best place to donate to, earning to give would be a more efficient use of time. However, I don't feel too confident in that. I don't know enough about how effective professionals actually are in practice, and LessWrongers in general seem to be extremely capable, even when venturing outside of their areas of expertise.
I also think a top-level post making the case for or against focusing on COVID vs. (other?) X-risk is a great idea.
I agree. I just posted this question.
Also see https://blog.givewell.org/2020/03/10/march-2020-open-thread/.
Some coronavirus-related problems are more tractable today than normal problems.
Hm, that's true, right now is a particularly good time to work on the coronavirus. I'm not sure if that outweighs the fact that other issues like xrisk are way bigger than the coronavirus though.
I totally agree that MMM got quite a few things wrong in that post. It's caused me to decrease my confidence in him too. However, the lifestyle factors point didn't strike me as one of those. We certainly shouldn't take his word for it, but it seems worth considering the question.
Those are some good points you make, thank you. I agree that there is something to be said about how "lifestyle factors" are a conscious choice that people made. To me that nudges things somewhat, but isn't a game changer. I don't think it makes it 10x less bad or anything.
The economic impact is a point that I think is crucial to the question of how bad this really is, and I think it's related to the questions I pose about how bad is it really to have less money. If bad economic impact means lower standard of living, and lower standard of living isn't really that impactful on happiness, then maybe bad economic impact isn't that bad. But I suspect that there are things I'm overlooking, and that bad economic impact is in fact relatively bad. So then, I update my viewpoint to being that it's a notable amount worse than lifestyle factors deaths, but still in the same ballpark, not 10x worse. My confidence in the "how bad is it to have a bad economic impact" question is pretty wide though, because it's not something I know much about.
with an imminent risk that 1-10% of everyone dies within the next two years.
Is that really a possibility? I imagine that governments would impose a strict quarantine before letting it get that bad.
For my low-income friends though, yes. Yes it is going to be that bad. Sometimes people don't have jobs. Sometimes people don't have savings. A large portion of people live paycheck to paycheck.
In the situation where you don't have savings or a job, here is what I'm imagining. The majority would have family or a friend they could stay with until they get back on their feet, which doesn't seem that bad. For those who don't have anyone to turn to, I assume homeless shelters would be an option, as opposed to literally dying on the streets without food, water or shelter. Homeless shelters do provide basic needs, so if you want to be really hardcore with the "happiness is all in your head" stuff, you should still in theory be ok. But I don't know much about what it's truly like; maybe there's more to it than that. On that note, to be clear, I don't mean to come across as insensitive or anything. I fully acknowledge that I might be wrong here. What I'm trying to do is explain what my model is and figure out where it might be wrong.
That's a great point, I totally agree. Like Isnasense mentioned in their comment, my faith in MMM is quite a bit lower after reading the post. However, it is still possible that correct about the points regarding how bad the coronavirus is compared to status quo thing like heart disease.
I don't think we should take his word for it, but thinking about it from first principles, it seems at least very plausible to me. But maybe I'm wrong, that's partly why I wrote this post. I'm curious to hear what others think.
To answer these questions it seems like it would be quite helpful to have domain specific expertise. So then, along the lines of comparative advantage, wouldn't it be more effective to earn to give? And following that thought, while the coronavirus is certainly scary, is it actually worth putting resources towards over things like existential risk reduction?
Perhaps the response to these points is that in practice, the coronavirus is particularly salient, and people are more likely to help out by doing research into these questions than they are with eg. existential risk reduction or earning to give.
Makes sense. So with that expectation of more money and less regulations, how does that affect our expectation that hospitals will be overwhelmed, and thus that death rates will be higher?
As things get out of hand, I would expect countries to throw a ton of money at it, basically like declaring war. When a ton of money is thrown at it, will hospitals still be overwhelmed?
You're saying that the post is interested in supporting defecting and causing societal harm for personal benefit?
I'm simply commenting on the "personal benefit" part without acknowledging the "good for society" part.
Not that the "good for society" part isn't important. Of course it is. But that doesn't mean we shouldn't have conversations about the "personal benefit" part in isolation.
I think that we should distinguish between two different questions. 1) Is what they're communicating good information for me? 2) Is what they're communicating good for society?
I interpreted "credibility" as related to the first question. And my point was that a) the lack of info and b) the lack of nuanced discussion of info makes me think that this credibility is reduced.
True. The downside would be that animal testing is slower, which is I think why jimrandomh was proposing human testing.
Yeah, if you could reduce the space of possible vaccines to a smaller set of plausible ones, that certainly makes sense.
This makes me wonder, why not just let people volunteer to test risky treatments in general? Because there'd be bad actors who try weird shit willy nilly and misrepresent it to people as more plausible than it really is, such that the harm done to people outweighs the advancements in knowledge? But what if you remove the profit motive and only give this power to government researchers? Would they have too many career-y incentives to be too aggressive?
But the space of possible vaccines is very large, I assume. So even with a ton of human testing that only takes three weeks, maybe this still doesn't help much?