I’m writing a book about epistemology. It’s about The Problem of the Criterion, why it’s important, and what it has to tell us about how we approach knowing the truth.
I’ve also written a lot about AI safety. Some of the more interesting stuff can be found at the site of my currently-dormant AI safety org, PAISRI.
I’m somewhat confused. I may not be reading the charts you included right, but it sort of looks to me like just rinsing with saline is useful, and that seems like it should be extremely safe and low risk and just about as effective as anything else. Thoughts?
I suppose you’d agree that there are in fact tradeoffs at play here and that the real question is what direction the scale tends to lean. And I suppose you are of the opinion that the scale tends to lean in favor of narrower, more targeted solutions than broader, more all-in-one solutions. Is all of that true? If so, would you mind elaborating more on why you are of that belief?
Scaling the business is different than getting started.
To get started it’s really useful to have a very specific problem you’re trying to solve. Provides focus and let’s you outperform on quality by narrowly addressing a single need better than anyone else can.
That is often the wedge to scale the business. You get in by solving a narrow, hard problem, then look for opportunities to expand your business by seeing what else your customers need or what else you could do given the position your in.
To give another example from a previous employer, Plaid got their start by providing an API to access banks, and they did everything they could to make it the best-in-class experience, with special attention on making the experience great for developers so they would advocate for paying a premium price over cheaper alternatives. That’s still the core business, but they’ve expanded into other adjacent products both as API access to banks has become easier to come by (in part thanks to Plaid’s success) and as customers have come looking for more all-in-one solutions to their fintech platform needs (e.g. a money-movement product so they don’t have to manage transfers on their own, alternative credit decisions tools, etc.).
Given your desire to do something that’s more lifestyle business than a high-growth startup, better examples might be to look at similar lifestyle products. In the LW-sphere there’s things like Complice and Roam, and outside LW you’ll find plenty of these that have been quite successful or were successful in the past (Basecamp is a prime example here, but I think Slack was arguably a lifestyle business that accidentally figured out how to take off when they pivoted to messaging away from MMOs, etc.).
Also matters what the experience is like. High prestige university allows you to get a job at a high prestige company. Low prestige university makes it a lot harder to get considered for jobs at high prestige firms. You’ll have to outperform high-prestige peers by, say, 50% to get noticed if you want access to the same sort of opportunities they get access to via prestige.
(To be clear, I’m not in favor of this sort of thing, I just want to be realistic about it and I wish someone had been real with me about it when I was 17 trying to decide where to go to college. Don’t rely on your ability to outperform others. Take every advantage you can get and then leverage them to do even more!)
And this really makes it hard for me as an “indie hacker” to do what people often recommend: solve one very specific problem. Find a niche. Something narrow and focused. “Zoom in”. This works in areas where problems have low cohesiveness, but not when they have high cohesiveness.
It’s really hard to solve a lot of problems well. The value of an all-in-one product is that you really don’t need anything else. Everything it doesn’t do, or doesn’t do well enough to meet your needs, is a ding against it, and it’s relatively easy to peel off specific problem spaces from the general problem of basic business operations needs.
Here’s a specific example from the company I’m part of (Anrok). We sell a product that aims to replace a SaaS business’s need for an accounting team to be compliant with sales tax & VAT. We compete against both all-in-one solutions from billing systems that offer to also solve this problem because the problem is hard and they are trying to solve at least two problems: billing and tax. Since we just try to solve tax, we can do a better job of it. We also compete against accounting firms who try to offer you all-in-one services, but again same problem, plus the cost is a lot higher because they use humans to do what we do with code.
So I wouldn’t worry too much about the existence of everything apps. It takes a long time to build a good everything app that actually does “everything” within its domain (think Salesforce or Jira), and even then “everything” is often achieved via outsourcing some of the everything to third parties who build plugins and integrations.
“Unconditional” makes a lot more sense if we think of it as “unconditional, conditional on my ability to think of conditions on my love”. This is what I think most people mean by unconditional love: they can’t think of any reasonable conditions on their love, and would discount unreasonable conditions as unusual or outside the realm of what’s meant by “unconditional”.
This is probably something like a shape-rotator vs. wordcel thing: shape-rotators take words literally and are uncomfortable with a word like “unconditional” unless there are literally no conditions, while wordcels are happy to say “unconditional” if the conditions are outside their Overton Window for reasons they would stop loving someone.
Unless you are going to one of the big prestige universities, I don’t think it matters which you choose all that much. Save money.
My experience is that this is right. The list of top-tier global institutions, in terms of prestige, is short: Oxford, Cambridge, Harvard, MIT, CalTech, maybe Berkeley, maybe Stanford and Waterloo if you want to work in tech, maybe another Ivy if you want to do something non-tech. The prestige bump falls off fast as you move further down the list. Lots of universities have local prestige but it gets lost as you talk to people with less context.
Prestige mostly matters if you want to do something that requires it as the cost of entry. If you can get in, it doesn’t hurt to have the prestige of a top-tier institution, but there’s lots of things you might do where the prestige will be wasted.
Sadly, this is a tough thing to know whether you will need the prestige or not. You’ll have to make an expected value calculation against the cost and make the best choice you can to minimize the risk of regret.
Perhaps, but I also feel like this is a real misunderstanding of politics being the mind killer. Rationality is critically important in dealing with real world problems, and that includes problems that have become politicized. The important-to-me thing is that, at least here on Less Wrong, we stay focused, as much as possible, on questions of evidence and reasoning. Posts about whether Israel or Palestine is good/bad should be off limits, but posts about whether Israel or Palestine are making errors in their reporting of facts in ways that can be sussed out using statistical analysis feel very much on brand.
For comparison, COVID was a hot button issue for a long time, and Less Wrong hosted tons of great posts about various mechanical things about COVID while avoiding many of the political issues. Less Wrong has also stayed away from topics like abortion and racism because there’s little to say on the topic that isn’t a thinly veiled attempt to argue over values. So while some aspects of the current Israel/Palestine conflict are fights over values and should be off limits here, I’d be pretty sad if we couldn’t talk about trying to understand the facts of the situation, like whether or not Palestinian death figures are correct, just like we’ve been able to talk about COVID origins and whether or not masks are effective and controlling the spread of COVID (if you can remember back to when that was a controversial topic!).
I actually don’t think the problem with this post is politics, but that it’s nothing more than a link post, and except in rare cases, I’d like to see people add something more than just provide links.
The analysis in the linked article itself is interesting and not obviously politicized (or at least, isn’t from a mistake-theory standpoint; it’s definitely political if you’re a conflict theorist, but then what isn’t!).
I don’t know if there’s data available to answer this question, but it’d be interesting to know if the relatively small middle class that existed prior to the 20th century also saw lower fertility than their lower and higher class peers. And in that case maybe falling fertility is a middle class phenomenon that ends when people exit the middle class either by making enough to no longer be economically middle class or by making little enough money that they cannot aspire to middle class dreams.
Here’s what Claude has to say:
There is some evidence that a similar fertility pattern existed in the past when the middle class was relatively smaller, particularly in the 19th century, but the data is more limited and the trends are not as clear-cut as in modern times.
In the 18th and 19th centuries, fertility rates tended to be higher among the aristocracy and wealthy classes, as well as among the poorest segments of society. The emerging middle class, which consisted of merchants, professionals, and skilled artisans, often exhibited lower fertility rates.
Here are some examples and observations from historical studies:
In England and Wales, during the late 18th and early 19th centuries, fertility rates were highest among the landed aristocracy and gentry, followed by the urban poor. The middle classes, such as merchants and professionals, had lower fertility rates.
In France, during the 19th century, fertility rates were lower among the urban middle classes than among the nobility and peasantry.
In the United States, during the late 19th century, fertility rates were higher among the wealthy and poor, while the emerging urban middle class had lower fertility rates.
However, it’s important to note that fertility patterns were influenced by a complex interplay of factors, including cultural norms, access to contraception, and economic conditions.
While the data from the 18th and 19th centuries is less comprehensive and more subject to regional variations, there is evidence that the middle classes, particularly in urban areas, were at the forefront of the fertility transition, with lower fertility rates than the wealthiest and poorest segments of society. This pattern is consistent with the idea that the marginal value of additional labor played a role in fertility decisions, although other factors were also at play.
I asked Claude for sources on the above claims. Number 2 might be hallucinated, but all the others I found quickly with Google:
Wrigley, E. A., Davies, R. S., Oeppen, J. E., & Schofield, R. S. (1997). English population history from family reconstitution 1580-1837. Cambridge University Press. This study analyzed parish records and found that in England, fertility was highest among the landed gentry and aristocracy in the 18th and early 19th centuries.
Livi-Bacci, M. (1986). Social-group fertility differential in France: An economic analysis. Population Studies, 40(2), 223-243. This paper examined data from the 1851 census in France and found lower fertility among the urban middle classes compared to the nobility and peasantry.
Haines, M. R. (1992). The New England Family Study: A study of fertility differentials among Pre-Industrial Urban and Rural Populations. Journal of Family History, 17(1), 17-35. This study analyzed fertility patterns in New England in the late 18th and early 19th centuries, showing lower fertility among the urban middle classes.
Hacker, J. D. (2003). Rethinking the “early” fertility decline in the United States: Integration of old evidence and new models. Social Science History, 27(4), 545-586. This paper argues that fertility decline in the United States began earlier than previously thought, with the urban middle classes leading the trend in the late 19th century.
Livi-Bacci, M. (1977). A history of Italian fertility during the last two centuries. Princeton University Press. This book examines fertility patterns in Italy and found that the middle classes were often at the forefront of fertility decline in the 19th century.
Who’s this ~no one? I came to see faith differently once I understood more of how the term (or another word in another language with the same base meaning of <trust>) is used in different spiritual traditions. Maybe few Christians and those primarily exposed to Christian memes conceive of faith in this way, but then this begs the question of why privilege their conceptualization of faith rather than looking for some some commonality between what people around the world seem to be pointing to when they say “faith” or a similar word to point to the idea of <trust> as part of a spiritual tradition?
I’ve always felt “predestination” is a funny way to phrase it. I get why it was chosen (the destination of your soul is already determined), but I think “deterministic” is better. It’s a really useful frame to understand that what will happen is what was already going to happen, and it’s just that you subjectively don’t yet know what the future will be and couldn’t compute it with certainty even if you tried (that’s why we give probability estimates for things!)
I’ve spent some time thinking about this and can share some thoughts.
I find the framing around preserving boundaries a bit odd. Seems like lack of preservation is one way things could be bad, but I don’t think it’s a full accounting for badness (or at least it seems that way based on my understanding of boundaries).
In humans, I strongly suspect that we can model badness as negative valence (@Steven Byrnes recent series on valence is a good reference), and that the reason “bad” is a simple and fundamental word in English and most languages is because it’s basic to the way our minds work: bad is approximately stuff we don’t like, and good is stuff we do like, where liking is a function of how much it makes the world the way we want it to be, and wanting is a kind of expectation about future observations.
I also think we can generalize badness from humans and other animals with valence-oriented brains by using the language of control theory. There, we might classify sensor readings as bad if they signal movement away from rather than towards a goal. And since we can model living things as a complex network of layered negative feedback circuits, this suggests that anything is bad if it works against achieving a system’s purpose.
(I have a bit more of my thoughts on this in a draft book chapter, but I was not specifically trying to address this question so you might need to read between the lines a bit.)
Goodness, in these models, is simply the reverse of badness: positive valence things are good, as are sensor readings that signal a goal is being achieved.
There are some interesting caveats around what happens when you get multiple layers in the system that contradict each other, like if smoking a cigarette feels good but we know it’s bad for us, but the basic point stands.
This statement would be false is, for example, we discovered that people could change other’s perceptions of the world by expecting them to be different and taking no other action.
Exactly. In each moment, the world is exactly as it is and can’t be anyway other than how we find it. Then it’s the next moment and the world is no longer the same as it was the moment before, yet is still however it is in that moment and no other way. The world is constantly changing from moment to moment, but always changing into exactly what it is.
I reject the claim that faith implies the world cannot change
Me too, which is why I didn’t write this.
Level Up told people to “Let your faith die,” and then contrasted faith with wonder.
I really don’t think that most people experience faith as being in opposition to wonder. It also suggests that faith is incompatible with the sort of progress the rationalist community wants.
This bit about faith points to something that frequently annoys me when interacting with my fellow rationalists.
I’m a person with faith now, but I wasn’t always, and it took me a long time to figure out what faith really means. I spent most of my life deeply misunderstanding what faith is because the Christians I grew up around often conflated faith with unwavering and unquestioning belief in dogma, to the point that even now it’s unclear to me if they meant anything else by the word. I only came around on faith once I realized it was just Latin for trust, and specifically trust in the world to be just as it is.
If a person is a Christian (especially a Nicene Christian), faith will include a metaphysical belief in God because they believe the world is God’s creation and evidence for God exists everywhere within it. I’m not a Christian and so don’t hold such a belief, but I can nevertheless have faith that the world will always be exactly as it is, and find refuge in my trust that I cannot be wrong about my experience of it prior to interpreting and judging it.
But lots of rationalists I know don’t get this. Like they can say the worlds “it all adds up to normality” but then constantly say and do things that suggest to me that they actually think that if they just try a little harder they might understand things well enough to remake the world in some fundamental way. They lack faith, and by extension humility. And while it’s good to see pain in the world and want to heal it, such healing will always be limited in effectiveness so long as the world is not seen clearly, and it’s my strong belief that someone is not seeing clearly if they don’t have faith in the world to be as it is.
Sorry for this ranty tangent, but the song line is plucking at a thread that I think is worth pulling, and I’ve not spent enough time writing about my thoughts here. Hopefully this is somewhat understandable.
In a sense, yes.
Although in control theory open- vs. closed-loop is a binary feature of a system, there’s a sense in which some systems are more closed than others because more information is fed back as input and that information is used more extensively. Memoryless LLMs have a lesser capacity to respond to feedback, which I think makes them safer because it reduces their opportunities to behave in unexpected ways outside the training distribution.
This is a place where making the simple open vs. closed distinction becomes less useful because we have to get into the implementation details to actually understand what an AI does. Nevertheless, I’d suggest that if we had a policy of minimizing the amount of feedback AI is allowed to respond to, this would make AI marginally safer for us to build.
Most AI safety policy proposals are of this type. I’m not suggesting this as a solution to get safe AI, but as a policy that government may implement to intentionally reduce capabilities to make us marginally safer.
Author’s note: This chapter took a really long time to write. Unlike previous chapters in the book, this one covers a lot more stuff in less detail, but I still needed to get the details right, so it took a long time to both figure out what I really wanted to say and to make sure I wasn’t saying things that I wouldn’t upon reflection regret having said because they were based on facts that I don’t believe or I had simply gotten wrong.
It’s likely still not the best version of this chapter it could be, but at this point I think I’ve made all the key points I wanted to make here, so I’m publishing the draft now and expect this one to need a lot of love from an editor later on.