Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Going Crazy and Getting Better Again, published by Evenstar on July 2, 2023 on LessWrong.

[Epistemic status: Processing after a long period of work and recovery, which is not yet over]

Background

Four years ago, I was in university. I was a casual rationalist, having been introduced by HPMOR, been active on rationalist tumblr, and then joined a loosely allied fiction writing community known as Glowfic. Things seemed pretty okay. I had known for a while that I was not neurotypical, having been diagnosed with ADD in high school and also having been in a gifted program.

I was also familiar, casually, with weird things happening to me and people around me. I was active in the hypnosis scene recreationally, and was fascinated by how people transformed under the influence of trance. I am also trans, and thus had experienced a major identity shift from my "cis male" self to a transfeminine identity.

That feminine identity was based on a female roleplay character that I had played as obsessively for several years - not an uncommon occurrence in trans people. She talked to me in my head - not an uncommon occurrence for authors. Eventually I decided I'd be her because she was cooler than my (dysphoric, beaten-down) self. And so we went on in a new identity, changing from our deadname to our new name. There was an attempt at coming out which... did not go well. I'll spare the details. I ended up closeted and under great stress to finish my degree.

I reached out to writing as my source of comfort, as I always had. I poured my hurt out into a newly-turned girl being abused by her vampiric sire. I started getting feelings and emotions from her - a sense of attraction to gothic and old-style clothing, a distinct fondness for lists and checkboxes. It was useful to me to like lists and checkboxes - after all, I was failing my degree from ADD - so I encouraged the persona, using a half-understood concept of Internal Family Systems and my experience with altered states achieved under hypnosis to try and develop the identity.

It worked! And we remain plural to this day, with a dozen or so other identities we've developed over time. We use these identities as different lenses to view the world through, ways of helping us not get too attached to any particular identity. We took the prescription to be a fox and not a hedgehog very seriously to heart. As far as we can tell this is both stable and useful; it's not what this post is about. It's simply necessary background so you can understand the level of weirdness in our life at this time.

So, here we were, trying to succeed at university, feeling like we didn't have enough willpower to manage to apply ourselves to our studies. And we had previous experience with Applying Weird Shit To Get Things Done.

So when we heard about Mastering The Core Teachings of The Buddha (which I will refrain from linking) we were interested. There was supposedly a way to gain better control over your emotions and thoughts that had been theoretically refined for a long period of time and had been studied at least a little by mainstream science! Medication didn't seem to work, so with a can-do attitude we set out to do some brainhacking.

In short, we got got.

The Psychosis

It is somewhat unclear, looking back now four years after my admission to hospital for acute psychosis, whether the meditation practices caused the altered states or whether the altered states caused the meditation practices. Certainly I had experienced many strange things in my life previously that are easily attributed to a schizotypal brain. Even before I picked up a religious bent and started fixating on ritual actions, I was having mood swings and feelings of distrust. What I can say for certain is that the ritual actions gave an excuse for me to experience all kinds of strange things, an...

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: VC Theory Overview, published by Joar Skalse on July 2, 2023 on The AI Alignment Forum.

In this post, I will give a brief overview of VC theory and computational learning theory. The reason for this is that I will later write a post about singular learning theory, in which I will refer back to this post. However, this post is self-contained, and may be interesting for anyone interested in theoretical machine learning.

What is Computational Learning Theory?

In short, computational learning theory (CLT) is what happens when you combine complexity theory with machine learning. Just like complexity theory is concerned with understanding what kinds of problems are hard to compute, or easy to compute, CLT is concerned with understanding what kinds of things are easy to learn, or hard to learn. This means that CLT seeks to identify classes of problems that you can learn with a small amount of data, and classes of problems that no algorithm can learn unless it gets a large amount of data. This also means that CLT produces generalisation bounds for different kinds of learning algorithms, and other such results.

Probably Approximately Correct Learning

In order to study what kinds of problems are hard to learn, and easy to learn, we first need a mathematical formalism that captures what it means to "learn a problem". The most popular model in CLT is given by the probably approximately correct (PAC) learning framework. This is a model of supervised learning for classification. It uses the following components:

We have an instance space X, which contains all the data that we might observe. For example, X may be a vector space, or it may be the set of all bit strings, etc. If we are doing image classification, then X may be the set of all RGB images of a certain size, and so on.

We have a set of concepts C, where every element of C is a function c : X -> {0,1}. This is the set of functions that we might be trying to learn. For example, if X is a vector space, then C could be the set of all linear classifiers on X. Alternatively, if X is {0,1}^n, then C could be the set of all Boolean functions over n variables with circuit complexity =< m. And so on.

We assume that we have a (potentially unknown) distribution D over X, and that there is an unknown function c : X -> {0,1} that we are trying to learn. To do this, our learning algorithm will observe a number of data points (x,c(x)), where each x is sampled iid from D. Based on this dataset, it will guess a function h : X -> {0,1}. We want this function to be similar to c with high probability.

Specifically, given a distribution D over X, a concept c \in C, and a hypothesis h : X -> {0,1}, we say that the error of h is P(h(x) != c(x)), where x is sampled from D.

A learning algorithm L is an algorithm that takes as input a dataset {(x, c(x))} of points, and returns a function h : X -> {0,1}.

We say that L is a PAC-learning algorithm for C if there exists a polynomial p, such that for all ϵ and δ, and all distributions D over X and all concepts c in C, if L is given access to at least p(1/ϵ,1/δ) samples from D, together with their labels according to c, then L will with probability at least 1−δ learn a function whose error is at most ϵ. Moreover, if there exists a PAC-learning algorithm for C, then we say that C is PAC-learnable.

In other words, PAC-learning is a mathematical formalism that describes supervised learning of classification algorithms. C is a set of classification functions, that we may be trying to learn. We want to know if there exists a supervised learning algorithm L, such that if L is given access to a small (polynomial) amount of training data, then it will with high probability learn a function with low error. We want this bound to hold for all c in C, and all D over X, so that we can be sure that this bound holds e...

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Why it's so hard to talk about Consciousness, published by Rafael Harth on July 2, 2023 on LessWrong.

[Thanks to Charlie Steiner, Richard Kennaway, and Said Achmiz for helpful discussion.]

[Epistemic status: my best guess after having read a lot about the topic, including all LW posts and comment sections with the consciousness tag]

There's a common pattern in online debates about consciousness. It looks something like this:

One person will try to communicate a belief or idea to someone else, but they cannot get through no matter how hard they try. Here's a made-up example:

"It's obvious that consciousness exists."

Yes, it sure looks like the brain is doing a lot of non-parallel processing that involves several spatially distributed brain areas at once, so

"I'm not just talking about the computational process. I mean qualia obviously exists."

Define qualia.

"You can't define qualia; it's a primitive. But you know what I mean."

I don't. How could I if you can't define it?

"I mean that there clearly is some non-material experience stuff!"

Non-material, as in defying the laws of physics? In that case, I do get it, and I super don't

"It's perfectly compatible with the laws of physics."

Then I don't know what you mean.

"I mean that there's clearly some experiential stuff accompanying the physical process."

I don't know what that means.

"Do you have experience or not?"

I have internal representations, and I can access them to some degree. It's up to you to tell me if that's experience or not.

"Okay, look. You can conceptually separate the information content from how it feels to have that content. Not physically separate them, perhaps, but conceptually. The what-it-feels-like part is qualia. So do you have that or not?"

I don't know what that means, so I don't know. As I said, I have internal representations, but I don't think there's anything in addition to those representations, and I'm not sure what that would even mean.

and so on. The conversation can also get ugly, with boldface author accusing quotation author of being unscientific and/or quotation author accusing boldface author of being willfully obtuse.

On LessWrong, people are arguably pretty good at not talking past each other, but the pattern above still happens. So what's going on?

The Two Intuition Clusters

The basic model I'm proposing is that core intuitions about consciousness tend to cluster into two camps, with most miscommunication being the result of someone failing to communicate with the other camp. For this post, we'll call the camp of boldface author Camp #1 and the camp of quotation author Camp #2.

Characteristics

Camp #1 tends to think of consciousness as a non-special high-level phenomenon. Solving consciousness is then tantamount to solving the Meta-Problem of consciousness, which is to explain why we think/claim to have consciousness. In other words, once we've explained why people keep uttering the sounds kon-shush-nuhs, we've explained all the hard observable facts, and the idea that there's anything else seems dangerously speculative/unscientific. No complicated metaphysics is required for this approach.

Conversely, Camp #2 is convinced that there is an experience thing that exists in a fundamental way. There's no agreement on what this thing is – theories range anywhere from hardcore physicalist accounts to substance dualists that postulate causally active non-material stuff – but they all agree that there is something that needs explaining. Also, getting your metaphysics right is probably a part of making progress.

The camps are ubiquitous; once you have the concept, you will see it everywhere consciousness is discussed. Even single comments often betray allegiance to one camp or the other. Apparent exceptions are usually from people who are well-read on the subject and may have optimized...

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Sources of evidence in Alignment, published by Martín Soto on July 2, 2023 on The AI Alignment Forum.

Summary: A short epistemological study to discover which sources of evidence can inform our predictions of action-relevant quantities in alignment.

This post follows Quantitative cruxes, although reading that first is mostly not required. Work done during my last two weeks of SERI MATS 3.1.

Sources of evidence

No researcher in any field ever makes explicit all of their sources of evidence. Let alone in a field as chaotic and uncertain as ML, in which hardly-earned experience and intuitions play a central role in stirring the tensor pile. And even less in a field with as many varied opinions and confusing questions as alignment. Nonetheless, even when researchers are just “grokking some deeper hard-to-transmit structure from familiar theory and evidence”, they need to get their bits of information from somewhere. Knowledge doesn’t come for free, they need entanglement with observed parts of reality.

Getting a better picture of where we are and could be looking, brain-storming or deepening existing sources, and understanding methodology limitations (as Adam Shimi’s efforts already pursue) can dissolve confusions, speed progress forward, help us calibrate and build common knowledge.

In reality, the following sources of evidence motivating any belief are way less separable than the below text might make it seem. Nonetheless, isolating them yields more conceptual clarity and is the first step for analysis.

1. Qualitative arguments

One obvious, theoretical source, and the most used in this community by far. The central shortcoming is that their abstractions are flexibly explanatory exactly because they abstract away detail, and thus provide more information about the existence of algorithms or dynamics, than about some relevant related quantities like how prevalent they actually are in a certain space, when do these dynamics actually start to appear and with how much steering power, etc.

Sometimes a tacit assumption might seem to be made: there are so many qualitative arguments for the appearance of these dynamics (and so few for the appearance of rectifying dynamics), that surely one of them will be present to a relevant degree, and early on enough! This seems like a sort of presumption of independence about yet unobserved structure: a priori, we have no reason to believe any one of these qualitative arguments have higher or lower quantitative impact, so we should settle on the vague prior of them all having similar effects (and so, the side with more qualitative arguments wins). While this is truly the best we can do when further evidence isn’t available, it seems like an especially fragile prior, ignoring the many possible interdependencies among some qualitative arguments (how their validity cluster across different worlds), and possible correlated failures / decoupling of abstractions from reality, or systemic biases in the search for qualitative arguments.

Incorporating some of these considerations is already enough to both slightly better inform our estimates, and especially better calibrate our credences and uncertainty.

Of course, usually qualitative arguments are informed by and supplemented with other sources of evidence that can provide a quantitative estimate, and then the presumption of independence is applied after incorporating these different sources (which is usually a considerably better situation to be in, unequivocally more grounded in reality). We can even explicitly reason qualitatively about the different relative magnitudes of some dynamics, as in How likely is deceptive alignment?.

And sometimes, in even less explicit ways, intuitive assessments of the strength of quantitative effects (or even the fundamental shape of the qualitative arguments) ar...

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Quantitative cruxes in Alignment, published by Martín Soto on July 2, 2023 on The AI Alignment Forum.

Summary: Sometimes researchers talk past each other because their intuitions disagree on underlying quantitative variables. Getting better estimates of these variables is action-guiding. This is an epistemological study to bring them to the surface, as well as discover which sources of evidence can inform our predictions, and which methodologies researchers can and do use to deal with them.

This post covers some discussion of quantitative cruxes. The second one covers sources of evidence (a topic interesting on its own). The whole write-up can also be read in this document. It will be expanded with a benchmark of thought experiments in alignment.

Work done during my last two weeks of SERI MATS 3.1. Thanks to Vivek Hebbar, Filip Sondej, Lawrence Chan and Joe Benton for related discussions.

Many qualitative arguments urge worries about AI. Most of these argue shortly and easily for the possibility (rather than necessity) of AI catastrophe. Indeed, that’s how I usually present the case to my acquaintances: most people aren’t too confident about how the future will go, but if some yet unobserved variables turn out to not line up well enough, we’ll be screwed. A qualitative argument for non-negligible chance of catastrophe is enough to motivate action.

But mere possibilities aren’t enough to efficiently steer the future into safety: we need probabilities. And these are informed by quantitative estimates related to training dynamics, generalization capacity, biases of training data, efficiency of learned circuits, and other properties of parameter space. For example, a higher credence that training runs on task A with more than X parameters have a high chance of catastrophe would make us more willing to risk shooting for a pivotal act. Or, a lower credence on that would advise continuing to scale up current efforts. Marginal additional clarity on these predictions is action-relevant.

Indeed, the community has developed strong qualitative arguments for why dangerous dynamics might appear in big or capable enough systems (the limit of arbitrary intelligence). But it’s way harder to quantitatively assess when, and in what more concrete shape, they could actually arise, and this dimension is sometimes ignored, leading to communication failures. This is especially notable when researchers talk past each other because of different quantitative intuitions, without making explicit their sources of evidence or methodology to arrive at conclusions (although luckily some lengthier conversations have already been trying to partially rectify this).

Addressing these failures explicitly is especially valuable when doing pre-paradigmatic science action-relevant for the near future, and can help build common knowledge and inform strategies (for example, in compute governance). That’s why I’m excited to bring these cruxes to the surface and study how we do and should address them. It’s useful to keep a record of where our map is less accurate.

Methodology

We start with some general and vague action-relevant questions, for example:

A. How easy is it to learn a deceptive agent in a training run?

B. How safe is it to use AIs for alignment research?

C. How much alignment failure is necessary to destabilize society?

(Of course, these questions are not independent.)

We then try and concretize these vague questions into more quantitative cruxes that would help answer them. For example, a concrete quantity we can ask about A is

A1. How small is the minimal description of a strong consequentialist?

(Of course, this question is not yet concrete enough, lacking disambiguations for “description” and “strong consequentialist”, and more context-setting.)

The ultimate gold standard for such concrete quant...

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Consider giving money to people, not projects or organizations, published by NinaR on July 2, 2023 on LessWrong.

When trying to improve the world via philanthropy, there are compelling reasons to focus on nurturing individual talent rather than supporting larger organizations, especially those with nebulous and unquantifiable goals.

Tyler Cowen's Emergent Ventures is a prime example of this approach, providing grants to individual entrepreneurs and thinkers who aim to make a significant societal impact. When asked how his approach to philanthropy differs from the Effective Altruist approach, Cowen answers:

I’m much more “person first.” I’m willing to consider, not any area—it ought to feel important—but I view it as more an investment in the person, and I have, I think, more faith that the person’s own understanding of what’s important will very often be better than mine. That would be the difference.

This model has been effective in the scientific community. Funding individual researchers rather than projects has been shown to foster more novel ideas and high-impact papers, emphasizing the value of the person-first approach.

The person-first approach is an effective diversification strategy. You are outsourcing the task of problem prioritization and strategy to highly competent individuals and trusting the result. This seems wise; I expect competence in executing effective solutions to problems to be highly correlated with competence in identifying important problems in the first place.

In the same way as angel investors can significantly influence the success trajectory of startups, investing in highly competent individuals early on can amplify their potential for making major progress. By observing their academic achievements or impressive abilities early on in life, you can often obtain meaningful evidence that someone can have a major positive impact.

Patronage is also a model well-suited to advancing many forms of creative or personal endeavor that promote a donor’s personal aesthetic or other hard-to-quantify terminal values. Investing in people can create new writing, music, art, and architecture in a more steerable way than generally giving money to these industries.

Why consider this model over donating to larger nonprofits where some employees will also be very talented? The answer lies in feedback loops and organizational efficiency. For-profit companies operate under tight feedback loops; they either provide value and thrive or fail to do so and perish. Nonprofits, however, especially those with hard-to-measure outcomes, lack these feedback mechanisms, making inefficiencies more likely. In larger organizations, many inefficiencies are amplified as coordination problems and operational overhead are more prevalent, wasting resources.

Another crucial variable in deciding whether to donate to larger organizations or lean towards a person-first approach is how measurable the outcomes you are looking for are. Some nonprofit organizations are doing significant legible and measurable good work, for instance, Against Malaria Foundation and the other GiveWell top charities, as well as quite plausibly the Bill & Melinda Gates Foundation. However, this is a small minority. In many cases, it is worth considering whether the problems nonprofits claim to tackle would be better tackled by funding competent individuals or for-profit organizations that can sustain themselves on the open market.

Why should we not assume that talented people will receive enough funding and support to do important good stuff as it is?

Indeed, capitalism provides an inherent mechanism for preference and information aggregation. By default, it is reasonable to assume markets are a reasonable source of truth regarding what humans want and need. However, free markets are not the silver bullet fo...

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: How Smart Are Humans?, published by Joar Skalse on July 2, 2023 on The AI Alignment Forum.

Epistemic status: free speculation

How intelligent should we expect AI to be, compared to humans, and how quickly should we expect it to reach this level of intelligence? This is of course an important strategic question, which has a large impact on our AI threat models. At the moment, the most common answers to these questions are "much more intelligent", and "very quickly". In this post, I will describe an arguably realistic scenario in which this would not be the case.

First of all, why should we expect AI to end up much more intelligent than humans? A common argument goes roughly like this: humans are much, much smarter than animals. We can build computers, and send things to space, but they can't. This proves that the range of possible intelligence is very wide. Moreover, it would be prima facie very surprising if human intelligence is at the top of this range. Therefore, we should expect it to be possible for AI systems to get much smarter than humans. Moreover, there is no reason to think that AI progress would slow down around human intelligence in particular. Therefore, we should expect AI intelligence to quickly far outstrip our intelligence.

This argument relies very crucially on the assumption that humans are much smarter than animals. But is this actually true? I'm not entirely convinced. First of all, there have been "feral" humans that grew up surrounded by animals. As far as I know, these humans are not obviously much more intelligent than animals (in terms of their ability to solve problems). This already casts some doubt on the notion that humans are much, much smarter than animals.

It is important to remember that humans, unlike all other species, are able to use complex language. This is a huge confounding factor, when we try to compare the intelligence of humans and animals. It is obviously very powerful to be able to exchange complex ideas, and build up knowledge intergenerationally. This would probably be enough to give humans a very large advantage, even if our intelligence was otherwise exactly the same as that of other primates.

Therefore, consider the following hypothesis: humans have an innate ability to use complex, combinatorial language, but all other species lack this ability (in the way Noam Chomsky thinks). In addition to this, humans are also somewhat (but not hugely) more intelligent than other primates (eg, 1.5x as intelligent, say).

As far as I can tell, this hypothesis roughly fits all our observations (?). However, if it is the case that the difference between humans and monkeys is mostly due to a one-shot discrete difference (ie language), then this cannot necessarily be repeated to get a similar gain in intelligence a second time. If that is the case, then we should perhaps expect AI to still end up quite a bit more intelligent than humans, but perhaps not to an incomprehensible extent (ie, we end up with AI geniuses, but not AI gods).

There are obviously a lot of ways that this conclusion could be wrong, and a lot of counter-arguments one could offer (eg, in most board games, AI has quickly gone from below human performance, to far above human performance, yadda yadda). However, I don't know of any knock-down arguments, and so I put some weight on something like this being true.

Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Bill Gates' 400 million dollar bet - First Tuberculosis Vaccine in 100 years?, published by NickLaing on July 2, 2023 on The Effective Altruism Forum.

Last week a young man Onekalit turned up with a nasty cough to a health center we operate in a youth prison in Gulu, Northern Uganda. The dry cough wore him down for over a month, but last week he finally managed to cough a bit of sputum into a small plastic container. The incredible Gates Foundation funded GeneXpert test confimed our fears – Tuberculosis

But Onekalit will be OK – after 6-9 months of gruelling treatment, the TB will be cured. He will not become one of the 1.5 million people that TB kills every year, more than double that of malaria. After covid subsided TB has now regained the dubious honour of the world’s deadliest infectious disease.

The Gates Foundation helped bring the amazing GeneXpert diagnostic test to places like rural Uganda, but Bill and co. are now going a step further making their biggest ever 400 million dollar bet on a vaccine that initial trials show may be 50% effective in stopping TB progress from latent infection to deadly lung disease. The first new effective TB vaccine in over 100 years.Surprisingly this vaccine has been sitting around (in a form) doing nothing much for around 20 years. GlaxoSmithKline (GSK) bought the patent almost 20 years ago, before publishing a trial which showed it was actually quite good and could save millions of lives, then decided they couldn’t make money from the vaccine so shelved it... Crazy stuff.

Unfortunately, our economic system is not set up to bring a vaccine which could save hundreds of millions of future lives to market. Fortunately our economic system does allow people like Effective Altruists and Bill Gates to donate their own stacks of cash towards life saving endeavours that the market has failed to bring to fruition.

This mind bogglingly expensive 550million dollar trail is necessary because TB is a slow disease. Slow to divide, slow to spread, slow to treat. Tracking and follow up TB takes far longer than for your average infectious disease. For malaria, within a year we can start to see whether a vaccine works. For TB it will take at least 5 times as long – 5 years or more before we know whether we are onto a winner.

If the vaccine really is 50% effective, it could save around 10 million lives in the next 25 years, not to mention helping prevent the terrifying Antimicrobial Resistance (AMR) that TB has already partly achieved.

500 million died from Smallpox (“but not a single one more”) – over 1 billion have died from TB. We remain far a from “not a single one more” in the case of TB – but this could be a spectacular step in the right direction.

Not his real name

Although from an effective altruism perspective the suffering caused by malaria is worse, because malaria kills mostly young children, whereas TB kills people of all ages.

I’m not sure about this, please correct me if I’m wrong

Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Micro Habits that Improve One’s Day, published by silentbob on July 1, 2023 on LessWrong.

This is my fifth attempt at writing this post. I’m starting to think that I’ve already spent way too much time on this topic, which I’m convinced is valuable, but maybe not so valuable as to spend 20 hours perpetually rewriting a post about it. So obviously my solution is to rewrite it again, but this time in bullet points.

Here’s a tl;dr: There are some habits people can pick up that are very cheap, and may have positive effects, but these effects are too small to reliably notice consciously. Hence these habits are often neglected. In this post I argue to take some of these habits more seriously, and if they’re low-cost enough for you to implement, stick to them even absent of any feeling of them being useful.

One way to look at habits is to look at two axes: usefulness and effort

Both can be positive or negative

“Good habits” typically are beneficial, but it takes some effort to install them

“Bad habits” are the opposite in both directions, they are detrimental in some important way, but it is more effortful to get rid of the habit than to stick to it

There are of course also things that are either both beneficial and effortless (like breathing), or detrimental and effortful (like banging your head against a wall), but we typically don’t think much about these two quadrants because there’s no reason to override our natural inclinations

There’s a particular area in the “good but effortful” space that I call “useful micro habits”:

Interventions that are beneficial, but also take very little effort to maintain

A common problem is that for some of them their usefulness lies below the threshold of perception: it’s hard to tell if they really do anything, because benefits are small and/or indirect

So people may try such habits for a while, and then often drop them again for the apparent lack of benefits

Some examples of such useful micro habits (note that both usefulness and effort differ between people, so not all these examples may fit the definition above for you):

Paying attention to better hydration

Keeping CO2 levels low in your home/office

Dressing well (even at home)

Supplementing creatine

Planning one’s day in advance

Exposure to sunlight early in the day

Taking regular short breaks

Breathing exercises & meditation

Leaving the house daily, even when it’s not necessary

Journaling

Hugs

(Partially) cold showers

Many of these things may provide, say, a marginal 0-10% improvement in productivity, happiness, health, or some other desirable quality

For any such micro habit, the positive effects may be lower than typical daily variance, which makes it very difficult to get a convincing feeling of the habit “working”

People relying mostly on such subjective impressions will dismiss such habits and not reap the benefits

What’s the evidence for such interventions being beneficial?

Note that I’m mainly trying to make the point that such interventions exist at all (even if they differ between individuals), and that it’s sometimes a mistake to expect beneficial habits to feel beneficial (particularly if the habits are very cheap to maintain)

So the post is not hinging so much on the particular examples I provided above; that being said, here's a small selection: the evidence for hydration being important (and many people drinking suboptimally little water) generally seems strong, increased CO2 levels negatively impact productivity, creatine has physical benefits and possibly cognitive ones, natural (early) light supports the circadian rhythm (and thus wakefulness and sleep quality), and cold showers are shown to increase dopamine levels

What do we do with that?

Batch several such habits together to make effects easier to recognize, instead of judging any micro habit in isolation

E.g. Mak...

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Forum Karma: view stats and find highly-rated comments for any LW user, published by Max H on July 1, 2023 on LessWrong.

I've been posting and commenting pretty frequently over the last few months, and I was curious about some stats. What started as a few GraphQL queries and some Python scripting turned into an interactive web app:

Enter a username, and it will give you some stats and a graph, broken down by post and comment karma. You can use the slider to adjust the date range, and the stats are automatically recalculated for the selected time period.

Another feature is the list of "Gems" - comments with at least a few votes that have the highest net karma score, i.e. comments that received strong upvotes from high-karma users, and few downvotes. I found that this often gives a better sense of a user's best comments than just looking at the highest-scoring ones.

It can be useful for finding your own historical top comments, or to get a sense for the best contributions of other users, which are often buried in their comment history and hard to access directly.

Interesting example users:

Me!

It works for the EA forum too, though you have to specify a full profile URL to disambiguate:

://forum.effectivealtruism.org/users/lizka

://forum.effectivealtruism.org/users/holdenkarnofsky

Limitations and caveats:

The app works by querying the LW GraphQL API. It makes a couple of GraphQL queries to fetch data for up to 5000 comments and 5000 posts all at once. I only fetch the minimal fields needed to calculate the stats, not the actual contents of the posts or comments. (For top comments links and preview, the comment data is fetched individually as needed, on hover). I think this shouldn't place too much stress on the LW backend, but I can take it down if the developers or mods find it to be problematic.

Data and stats for users with more than 5000 comments will be incomplete. I tried for a while to find a GraphQL query that supports fetching old comments for users with lots of comments, but couldn't find one (offset doesn't work; the API returns an error for too-big offsets.) If any LW devs (or anyone else) finds a GraphQL query that allows fetching old comments for users with >5000 comments, I can fix this. GreaterWrong somehow accomplishes this to implement sort by top or old, but they appear to be doing it on the backend, so I couldn't reverse engineer it.

The comment previewer is pretty basic, since I'm just dumping the raw HTML from LW into a tooltip with minimal formatting. I might improve this at some point.

Stats for old users will be somewhat inaccurate - I assumed for the purposes of removing the user's self-vote (which doesn't count towards their karma) that everyone always strong upvotes their own posts and weak upvotes their own comments, and that strong / weak upvotes are worth the amounts described here. This will be slightly inaccurate for contributions from before strong voting was introduced, and more inaccurate for contributions from before self-voting was introduced. I was mainly interested in stats for more recent content, but if anyone has a handy schedule of when all the voting rules were implemented / changed, I can probably make it more accurate pretty easily.

Karma is counted as if it were all received at the time the post or comment was first posted. In my experience, the vast majority of votes on something happen relatively soon after it is posted - people tend not to vote on old stuff too much. So this introduces only a small inaccuracy.

Aside: this was the first nontrivial greenfield project that I've developed since GPT-4 was released. I used some technologies I was less familiar with (GraphQL, Chart.js, Tailwind CSS), and had GPT-4 write as much of the code as possible. I've been using GitHub Copilot for day-to-day programming for a while, but ...