Self-Embedded Agent's Shortform 2021-09-02T10:49:45.449Z
Are we prepared for Solar Storms? 2021-02-17T15:38:03.338Z
What's the evidence on falling testosteron and sperm counts in men? 2020-08-10T08:58:47.851Z
[Reference request] Can Love be Explained? 2020-07-07T10:09:17.508Z
What is the scientific status of 'Muscle Memory'? 2020-07-07T09:57:12.311Z
How credible is the theory that COVID19 escaped from a Wuhan Lab? 2020-04-03T06:47:08.646Z
The Intentional Agency Experiment 2018-07-10T20:32:20.512Z


Comment by Self-Embedded Agent on All Possible Views About Humanity's Future Are Wild · 2021-09-09T21:25:00.696Z · LW · GW

Fair enough. 

Comment by Self-Embedded Agent on All Possible Views About Humanity's Future Are Wild · 2021-09-08T14:08:15.065Z · LW · GW

We can't build Von Neumann probes in the real world - though we can in the digital world. 
What kind of significant (!) new information have we obtained about the feasibility of galaxywide colonization through Von Neumann probes?

Comment by Self-Embedded Agent on Self-Embedded Agent's Shortform · 2021-09-07T08:37:16.144Z · LW · GW

Failure of convergence to social optimum in high frequency trading with technological speed-up

Possible market failures in high-frequency trading are of course a hot topic recently with various widely published Flash Crashes. There has a loud call for a reign in of high frequency trading and several bodies are moving towards heavier regulation. But it is not immediately clear whether or not high-frequency trading firms are a net cost to society. For instance, it is sometimes argued that High-Frequency trading firms as simply very fast market makers. One would want a precise analytical argument for a market failure.

There are two features that make this kind of market failure work: the first is a first-mover advantage in arbitrage, the second is the possibility of high-frequency trading firms to invest in capital, technology, or labor that increases their effective trading speed.

The argument runs as follows.

Suppose we have a market without any fast traders. There are many arbitrage opportunities open for very fast traders. This inaccurate pricing inflicts a dead-weight loss D on total production P. The net production N equals P-D. Now a group of fast traders enter the market. At first they provide for arbitrage which gives more accurate pricing and net production rises to N=P. 

Fast traders gain control of a part of the total production S.  However there is a first-mover advantage in arbitrage so any firm will want to invest in technology, labor, capital that will speed up their ability to engage in arbitrage. This is a completely unbounded process, meaning that trading firms are incentived to trade faster and faster beyond what is beneficial to real production. There is a race to the bottom phenomenon. In the end a part A of S is invested in 'completely useless' technology, capital and labor. The new net production is N=P-A and the market does not achieve a locally maximal Pareto efficient outcome.

As an example suppose the real economy R consults market prices every minute. Trading firms invest in technology, labor and capital and eventually reach perfect arbitrage within one minute of any real market movement or consult (so this includes any new market information, consults by real firms etc). At this point the real economy R clearly benefits from more accurate pricing. But any one trading firm is incentivized to be faster than the competition. By investing in tech, capital, suppose trading firms can achieve perfect arbitrage within 10 microseconds of any real market movement. This clearly does not help the real economy R in achieving any higher production at all since it does not consult the market more than once every minute but there is a large attached cost.

Comment by Self-Embedded Agent on All Possible Views About Humanity's Future Are Wild · 2021-09-06T21:02:13.476Z · LW · GW

No. Not to take away Holden's thunder but Von Neumann already postulated the possibility of self-replicating probes colonizing the galaxy. Indeed, I might be wrong here but it is exactly this kind of thinking that drove Fermi to pose his famous question. 

Most of this stuff was fairly common in extropian/hard scifi Usenet circles in the 90's I believe. 

Comment by Self-Embedded Agent on Book review: Shut Out · 2021-09-06T15:22:15.365Z · LW · GW

Thank you for your comments. You are probably right that most economists would disagree with that sentence, but it was the sense I got from this particular economist - perhaps his views are more idiosyncratic than he presented them. I am not knowledgeable to say much more I'm afraid. 

Comment by Self-Embedded Agent on Book review: Shut Out · 2021-09-05T09:25:49.655Z · LW · GW

I think the conventional narrative under academic economists is that the failure of the Fed to bailout [even better: nationalize] Lehman Brothers which led to a liquidity crises. Since our modern economy is so tied up with a smooth-functioning financial sector this led to the Great Recession. The housing bubble was mostly a distraction that would have had no large impact if it wasn't for this mistake.

The take-away message [if this narrative is indeed correct]: the financial sector is really important, federal banks are unelected institutions where a single bad call can have enormous repercussions, people who 'predicted the financial crisis' are overrated - the failure to bailout Lehman Brothers was not seriously predicable in advance, narratives around economic fundamentals are wrong. 

source: talking with a mainstream economist I respect, Jacob Falkovich

I don't know enough about these things to have my own opinion, but I thought I might share this 'conventional' narrative and what I found surprising about the implications. 

Comment by Self-Embedded Agent on capybaralet's Shortform · 2021-09-03T10:38:51.026Z · LW · GW

Strong agree.

Comment by Self-Embedded Agent on Self-Embedded Agent's Shortform · 2021-09-02T11:50:34.012Z · LW · GW

Measuring the information-theoretic optimizing power of evolutionary-like processes

Intelligent-Design advocates often argue that the extraordinary complexity that we see in the natural world cannot be explained simply by a 'random process' as natural selection, hence a designer. Counterarguments include:

  • (seemingly) Complicated and complex phenomena can often be created by very simple (mathematical) rules [e.g. Mandelbrott sets etc]
  • Our low-dimensional intuition may lead us astray when visualizing high-dimensional evolutionary landscapes: there is much more room for many long 'ridges' along which evolution can propagate a species.
  • Richard Dawkins has a story about how eyes evolved in step-like manner from very primitive photo-receptors etc. In this case he is mostly able to explain the exact path.

To my eyes these are good arguments but they certainly are not conclusive. Compare:

  • Stuart Kauffmann has a lot of work ( see 'At Home in The Universe') on which kind of landscapes allow for a viable 'evolution by natural selection'. Only some landscapes are suitable, with many being either 'too flat' or 'too mountainous'. Somewhat hilly is best. Most systems are also too chaotic to say anything useful
  • Population geneticists know many (quantatitive!) things about when different evolutionary forces (Drift, Mutation, Sexual Recombination) may overwhelm natural selection
  • Alan Grafen has a fairly precise mathematical framework that he says is able to determine when Darwinian-Wallacian evolution is maximizing a 'fitness function'. Importantly, not all situations/ecosystems can support the traditional 'survival of the fittest' interpretation of evolution

The take-away message is that we should be careful when we say that evolution explains biological complexity. This might certainly be true - but can we prove it?

Comment by Self-Embedded Agent on Self-Embedded Agent's Shortform · 2021-09-02T11:15:12.454Z · LW · GW

Three normative principles:

  • Ole Peters' Principle/Kelly betting: Given a wager pick the one that maximizes your growth rate. In many-cases this coincides with log-wealth.
  • Efficient Market hypothesis: when betting/trading in a market, we cannot multiply our initial capital  with a large fraction  over the risk-less rate .
  • Maximum Entropy Principle: given constraints  on an unknown probability probability distribution , your 'best' estimate is  that maximizes the entropy under the constraints .

How might these be related?

Comment by Self-Embedded Agent on Self-Embedded Agent's Shortform · 2021-09-02T11:10:55.823Z · LW · GW

[This argument is not complete]

Block Universe vs Evolution

I find the block universe esthetically displeasing. But is there a coherent counterargument I could make?

Suppose we have a population of Frogs. Some of the Frogs have stronger and longer legs and are able to jump further. Now we observe at a later time that that the ratio of Long-legged Frogs to the Short-legged Frogs has (significantly) increased. Naturally we would turn to an natural selection story: the ability to jump further increases the chance of survival for a Frog and meant that they the longer-legged Frogs have outcompeted the short-legged Frogs. However, there is of course also a possibility that this happened purely by chance -> Drift hypothesis. 

The block-universe seems at odds ABORT

  • I think this was already observed by Judea Pearl: Causal identification is tightly wrapt up with 'Flowing Time'.
Comment by Self-Embedded Agent on Self-Embedded Agent's Shortform · 2021-09-02T10:49:45.777Z · LW · GW

What determines the ratio of long-term vs short-term investment?

  • What determines the ratio of long-term vs short-term investment in an economy (surely this is known?)
  • Long-term investment can be thought of as a cooperation of your Current Self with your Future Self. Entities with low-time preference/Long-term investment cooperate more intertemporally.
  • Following up to the above: smart people have low-time preference and thtis effect is surprisingly large (think about Yudkowsky's $500 vs 15% chance for 1 million dollars having <p=0.005 significance correlation with 'cognitive reflection) . Smart people also cooperate much more, and this effect is also quite large (don't remember the exact details). -> Might these effects be related?
  • - Smartness might allow for more approximate Lobian cooperation - intelligence might lower randomness in expected social interaction -> in repeated situations/ situations that might be exponentially important. -> I forgot the namebut there is this theory that a small increase in fidelity in multi-step production steps is the difference between able to make a functional aeroplane and not being able to do so and that this underlies much of the very large GDP differences we see. This seems sensible and similar
  • Perhaps the same reasoning could be applied to low-time preference: we could see long-term investment as a many-step current self-future self cooperation. Increasing the fidelity/ cooperation probability at each step by a small amount may exponentially increase long-term investment/low-time preference.

Interesting but doesn't explain quite explain everything

What determines the ratio of K-selectors vs r-selectors?

  • There might be a group-selection vs individual selection story going on: our intertemporal Selves are different but highly similar individuals.
  • Highly similar individuals seem to be able to cooperate better. -> (approximate)Lobian cooperation?
  • Alternatively, we could tell a evolutionary story: self-replicating processes are something like attractors in the space of all processes?  [see also Block Universe vs Evolution/ autopoiesis below]
  • Risk- appetite might be a form of cooperation. Indeed 'Being able to choose a risky but high EV option over a sure pay-off' [Like 500vs 15% chance of 1 million] is like you cooperating with your Counterfactual Selves.
  • How does this interface with Kelly Betting/Ole Peters stuff?  Interestingly Risk-appetite and low-time preference seem to both be cooperation, but the first is 'space-like' while the other is 'time-like'.
  • How does this interface with Armstrong's Antropic Decision theory ? Armstrong emphasizes that the different answers in Sleeping-Beauty paradoxes arise from different cooperation/utility functions.
  • Classically we would say utility functions are god-given, but following Ole Peters we would like to pick out canonical utility functions in some way.
  • Here we get back again at group-selection vs individual selection. Here I am of course thinking of Okasha's Evolution and Levels of Selection
  • Aggregates that are more internally homogenous have higher 'meta-fitness': they are more able to compete  (Mein Kampf principle). See biological organism with good cancer suppression or indeed etnically and culturally homogenous societies militarily beating far larger empires - Turchin's Asabiyah [But! this might also be a purely statistical artifact: there probably are many more homogeneous societies than heterogeneous societies]
  • So for a gene $G$ there might be a tradeoff on different levels: on the first level it may pay-off to act selfishly, on higher levels it will pay off to cooperate. [of course this exactly what Okasha investigates]
  • How does idiosyncratic risk versus aggregrate risk play into this?
Comment by Self-Embedded Agent on Progress, Stagnation, & Collapse · 2021-07-28T13:48:26.353Z · LW · GW

Dear Abram,

I have a little bit of a knee-jerk reaction to 'complex, nuanced, multicausal' explanations by historians in general and the Late Bronze Age collapse specifically. On a reread my reaction was perhaps a little overzealous.

The system collapse theory, due to Joseph Tainter(?), is not crazy. It does suffer from being overly flexible, hard to test etc. 

 I agree that "I still think knowledge can be lost quickly in times of chaos, particularly when population takes a nose-dive."

The following is also plausible

" a good modification to the progress equation would be to lose progress if population dips below some number"

Something like this is often argued to have happened with Tasman island, with Tasmanians losing basic technology like fishing, firemaking, and generally being less technologically sophisticated than their ancestors 8000 years before. However, last time I actually checked the details; Tasmanians losing firemaking at least seemed to be a myth.

Two undeniable historical cases of losses of technological ability are the late Western Roman Empire and the end of the Hellenistic Age. This is probably not tied to large changes in population size. Knowledge being lost quickly in times of chaos is not implausible, e.g. the siege of Syracuse or the Crisis of the Third Century. Many theories about these historical epochs are possible and contested. 

EDIT: another case of a large loss of knowledge is the Sacking of Baghdad (1258) by the Mongols. It does fit in "I still think knowledge can be lost quickly in times of chaos, particularly when population takes a nose-dive." On the other hand, is no need for an overtly complex theory. Horse-archer based steppe peoples burning down settled-down societies is a repeating theme in history. 

EDIT2: I looked into Joseph Tainter's theory of system collapse. He looks at three cases: Chaco collapse, Maya collapse, fall of Rome. From my reading of the literature there is fairly good evidence for a drought explanation for the Maya collapse - actually there are multiple Mayan collapses, drought based stories are apparently plausible for all three of them, but I am not an expert. My impression was that most experts in the Chaco favor enviromental causes as well - I don't know enough to have a strong opinion. 

The fall of Rome is of course hotly contested, so I won't go into it. Suffice to say there are many possibilities here. A strong contender is probably Peter Turchin's Assabiyah theory. 

Comment by Self-Embedded Agent on Progress, Stagnation, & Collapse · 2021-07-25T13:16:42.332Z · LW · GW

Dear Abram,

The system collapse theory of Bronze Age Collapse is nuanced, complex, sophisticated and most likely wrong.

As a general rule, historians love explanations that emphasize 'complexity' and 'nuance'. Be very wary of these explanations. They sound great as a verbal explanation but they are highly flexible hypotheses that are difficult to falsify and can easily explain any amount of data. The perfect fit for verbal bullshitters - which probably accurately describes the vast majority of non-quantitative social theorists -like historians.

A little experience with dynamical systems will show how problematic these models are. Be wary anytime somebody is talking about 'multicausal' explanations. There are typically two regimes: something is made out of many many independent factors, giving Gaussian distributions or there is one or two dominant eigenfactors that explain the vast majority of the variance. I could say more about this. 

The most likely explanation for Bronze Age collapse is changes in warfare: elite corps of chariot archers losing out to javelin -throwing massed infantry consisting of a melange of experienced mercenaries. Earlier infantry-based tactics were ineffective against the sophisticated hit-and-run tactics of corps of chariot archers but eventually new tactics were devised - like using javileneers to hit the horses and wheels of chariots. 

 Writing and much of the archeological finds of this period rely on a demand for luxury- trade and elite scribes. Once the chariot-archer elites disappeared so did their luxury trade and elite scribes; hence a 'collapse'. 

Much more can be found in Drews' excellent book "The End of the Bronze Age: Changes in Warfare and the Catastrophe ca. 1200 B.C."

Comment by Self-Embedded Agent on How credible is the theory that COVID19 escaped from a Wuhan Lab? · 2021-07-09T12:12:01.618Z · LW · GW

I hesitate to give a probability.

Hindsight is 20/20 but the 85% probability was too high in a certain sense. 

I am talking about the following phenomenon: if you had forced me to bet a given fixed amount of money either way right at that moment I would take the 85% implied betting odds. However, if you had (i) offered me such a bet freely, (ii) allowed me to set the amount of money in the bet, (iii) and/or given me more time to research the issue I would ahve refused (i), only bet a small amount on (ii) and certainly spend more time researching the core claims (iii).

The above considerations show the problems with simply giving fixed probability distributions for belief. Certainly, the following phenomena should be considered:

  • There is a measure on how strongly a belief is held. The strength is often measured in the bet size.
  • A difference between belief and belief-in-belief
  • Somebody offering you a bet is implictly giving information about adversarial optimization (I once spoke with a guy in Finance that the bid-ask spread in markets is a measure of this)
  • There some sort of parameter in a reasoners brain that determines how long that reasoner will spent researching a given issue. Presumably this is set abnormally - perhaps dangerously so! - in the median rationalist.

I am sure there is a literature on this topic which I am now excited to check out! I am sure LessWrongers have thought about this... I would be grateful for any references!

Two important reasons I now favor natural causes: (i) most of the technical evidence seemed to have fallen through - it is always hard to ascertain the validity of technical evidence as a layman, but that was my impression. My initial update rested a lot on (a) not having much evidencemass either way so being easily convinced (b) the lableak piece by Nicholas Wade - even if some of the evidence holds up I feel it overstated its case/omitted contrary evidence. 

 (ii) the a priori probability of a lab leak origin of COVID-19 should be quite small; of all viri we are confident of a source it is natural, and though lab leaks have happened in the past, there are many more naturally- caused pandemics than human-caused pandemics


"Okay, that was all very very interesting... but please .... I insist"

You put a gun to my head

 "What are your bettings odds?" 

"Okay, okay 80% chance natural cause, 20% something else".

Comment by Self-Embedded Agent on How credible is the theory that COVID19 escaped from a Wuhan Lab? · 2021-07-06T12:30:47.830Z · LW · GW

Update: I have again updated away from the lab leak theory. Most of the evidence for a lab leak seems overstated after a closer look. See for instance this skeptical take by potholer54.

Comment by Self-Embedded Agent on Search-in-Territory vs Search-in-Map · 2021-06-11T22:19:25.524Z · LW · GW

A very basic yet, to my mind, novel and profound distinction. Thank you, John!

Comment by Self-Embedded Agent on Heliocentrism in the ancient era · 2021-06-04T13:20:34.082Z · LW · GW

I emailed Viktor Blasjo, an expert on the history of mathematics and physics on this post. I quote:

Very nice, but the author neglects another strong argument: how Aristarchus's own treatise on the sizes of the sun and the moon suggests physical support for heliocentrism. See pages 86-88 of

Comment by Self-Embedded Agent on We need a career path for invention · 2021-05-29T15:00:21.999Z · LW · GW

Let me outline a very simple model of technological progress. 

Innovations get exponentiably harder. As the lower-hanging fruits get picked, one needs to combine and master more and more previously understood scientific knowledge to get higher. Moreover, fruits higher up in knowledge tree may be intrinsically harder to pick. 

As an example, see the ever increasing material complexity and size of particle colliders. I.e. LHC to earlier backyard garage colliders. 

Our ability to pick fruit does still increase but that ability may not necessarily increase comparably fast.

Between ~1850-1950 the effective amount of brain power increased substantially. More important than just population growth is probably increased literacy, urbanization, formal education, improved nutrition, improved communication methods to facilliate knowledge etc etc. It might have increased the total effective amount of brain power applied by ~ two orders of magnitude.  

By comparison in the period ~1950-2020 the total amount of brain power may have only increased  ~a couple times. 

One has to be cognizant of the fact that innovations are made by a tiny percentage of highly excentric and talented individuals. This may not always line up with the mean of the population. 

We see substantially more collaboration in Science, and a much larger number of scientists, and a greatly increased amount of specialization. The sum of human scientific expertise does not fit comfortably in the skull of an unedited Homo Sapiens, and this capacity difference is increasing over time. 

Comment by Self-Embedded Agent on How credible is the theory that COVID19 escaped from a Wuhan Lab? · 2021-05-15T15:56:33.303Z · LW · GW

An update. 

This recent article by Nicolas Wade: is particularly damning, going into details into many lines of evidence all pointing towards COVID-19 accidently being released from the Wuhan Lab.  

I recommend anybody interested in reading the article. It has made me update substantially towards an accidental-lab-release origin for COVID-19.

My current estimate for such a lab-release origin is ~85%. 


Update: Some further googling brought up some skepticism towards some of Nicolas Wade's claims, here and here

my best estimate is now 65% for a lab-release. This is still a substantial update from my initial ~20% for a lab release origin for COVID-19 (estimated in hindsight).

Comment by Self-Embedded Agent on We need a career path for invention · 2021-04-30T13:20:17.719Z · LW · GW

Alternative theory: most of the low-hanging fruit has been picked. In the information-technology side of things we see an enormous amount of innovation, often because it is possible to create new products with a garageband skeleton crew. In the physical realm, thing are simply much harder, much more specialized. 

Also, there is a negative feedback loop: since the world is much more specialized there are fewer people working in technical fields, and hence fewer people having the basis in technical know-how that might lead to chance-inventions [which I learned from you was a typical story during the early Industrial Revolution in Great Britain]. 

How would you respond to this take?

Comment by Self-Embedded Agent on Beliefs as emotional strategies · 2021-04-11T07:37:58.347Z · LW · GW

I enjoyed this write-up. 

Buuut... I have to ask: how do we know your emotional stories are ' correct' and not just-so-stories?

Sure I can imagine that the whole story with the mother and the child is factually correct, yet is it really the reason that the child is giving up prematurely?

Maybe just maybe the child simply doesn't like doing hard things, doesn't like math, or is genuinely untalented or maybe there is completely different crazy complex story about the relations with their relatives, some deep early childhood trauma.

From personal and other people's experiences I agree that hearing the 'right emotional story' can be satisfying and give a sense of relief. Sometimes it will even lead me/other people to make short-term change. But that's different from it making any long-term changes (like changing one's work ethic), and even that doesn't mean it is true [just a helpful delusion!].  

Comment by Self-Embedded Agent on [Prediction] The Consequences of Radical Reform · 2021-03-11T15:29:55.981Z · LW · GW

Is it possible that political reform may have a causal influence on economic outcomes? Sure. 

Does this study prove and - more importantly- rigorously quantify this influence? No. 

The statistics are simply not powerful enough to yield such strong conclusions. 

The effect size is small, there is clear garden-of-forking-paths going on (only finding a real effect on urbanization), one has to mediate for a simple west-east axis. It is also highly suspicious that the effect only starts to take effect after 1850. etc etc etc

This could be pure noise. The study does not adequately reject the null hypothesis. 

Comment by Self-Embedded Agent on [Prediction] The Consequences of Radical Reform · 2021-03-10T11:12:52.540Z · LW · GW

I am quoting notJosephConrad :

Would it really kill economists to stick some reference to effect size in the abstracts of these things? If this were a clinical trial, and the authors were like "hey, good news, napoleolimumab increases modernization, now FDA-approved for your unsightly problems with Prussian stagnation" they'd have the common decency to say "10 years of french reforms increases the primary endpoint of urbanization in 1900 by 9% from 41% to 50%" But instead i had to hunt through the text for it like a goddamn animal (page 23, if you care. They also have a point estimate of 36% GDP increase.

But now that they forced me to go through the paper, some not-terribly informed thoughts:

1. I have the usual concerns about econ research -- were any of these analyses pre-specified? How many different analyses were tried before they went with this one, etc.

2. If I am reading this correctly, by 1850 no changes are seen. So all the positive effect of the new institutions is from 1850-1900. Interesting.

3. Riffing on '2' -- maybe this can be spun as another example of "industrialization changes everything" or "conservatism is a better default in the absence of massive scientific/technological change." Blowing up institutions in 700 AD does you no good, because there's no innovation to take advantage of, you just get chaos. Blowing up institutions in 1800 AD helps, because it enables social shifts to take advantage of new modes of production.

4. And just for honesty: my prediction was "can't discern an obvious effect" (which in retrospect was idiotic given that if it were a null effect it never would have been published)

In other words, this is chasing noise, zero-value research. Daron Amoglu is the most-cited economist in the world. His speciality is development economics. One of the foundational axioms is that historical contigencies have large effects on economic development decades or centuries later. This is a highly problematic assumption, one I think is mostly wrong. However, it is a treasure trove for telling 'interesting stories'. 

All these questions are implictly highly politicized. There is a pretension that economists are doing value-free research. This is a good ideal but very difficult; in practice there is an enormous incentive to p-hack noise into providing fodder for socially-approved stories.

My take-away message:

  1. You cannot trust 99% of social science research &Almost all social science researchers are highly sophisticated frauds 

 2. All big claims about the social world are so highly correlated with an implicit background worldview that is highly politicized. 

3. Never trust anything in a social science paper that you have not personally checked. 

Comment by Self-Embedded Agent on What's the evidence on falling testosteron and sperm counts in men? · 2021-02-28T12:12:08.698Z · LW · GW

Thank you Waveman.

For the record, it wasn't my intention to antagonize anybody with my tongue-in-cheek phrasing. Nostalgic macho cowboys have feelings too =)

From Waveman's article:

These findings indicate that the past 20 yr have seen substantial age-independent decreases in male serum T concentrations, decreases that do not appear to be the consequence of the contemporaneous trends in health and lifestyle considered here. It remains unclear to what these apparent population-level decreases in T are attributable.

Comment by Self-Embedded Agent on What's the evidence on falling testosteron and sperm counts in men? · 2021-02-28T12:07:06.354Z · LW · GW

Coming back at this issue it seems possibly quite serious, it certainly seems underexamined.

I found this Vox article and the wikipedia page useful. It seems quite clear that there is quite a large drop in Western countries. That a similar drop isn't seen in non-Western countries is especially telling.

The Vox article also points towards research into in-utero exposure to certain chemicals permanently decreasing sperm count. This also seems quite scary.

For example, there’s compelling data that in utero phthalate exposure is linked to a decrease in something called “anogenital distance,” or AGD, in male babies. AGD is the space between the anus and the genitals, and a man’s is usually twice as long as a woman’s. In men, a shorter AGD has been associated with poorer semen quality, less testosterone, and a higher risk of infertility.

One of the scientists to first describe this phenomenon, Mount Sinai’s Shanna Swan, told Vox that these early chemical exposures have lifelong results. “The lowered androgen and alteration of development that happens in utero [results in] changes that are lifelong,” she said. “They are not correctable. Maybe in the future with genetic modification, you can alter a man’s germ cells that produce his sperm — but for now, if those are impacted adversely, a man will have a lowered sperm count his whole life.”

Separately, a March 2019 study found exposure to phthalates and polychlorinated biphenyl — again, two ingredients found in plastics — damaged the quality of sperm exposed to the chemicals.


This Forbes article is a typical piece about falling testosterone counts in men. Much is made of changing cultural mores, social expectations of men, etc. The biological component seems to get mentioned only in passing.

I found this article by urologists quite frightening. The effect size is eye-popping.

After controlling for confounders—including year of study, age, race, BMI, comorbidity status, alcohol and smoking use, and level of physical activity—total testosterone was lower among men in the later (2011-2016) versus earlier (1999-2000) cycles (P < 0.001). Mean total testosterone decreased from 1999-2000 (605.39 ng/dL), 2003-2004 (567.44 ng/dL), 2011-2012 (424.96 ng/dL), 2013-2014 (431.76 ng/dL), and 2015-2016 (451.22 ng/dL; all P < .0001).

Comment by Self-Embedded Agent on Different kinds of language proficiency · 2021-02-26T19:03:05.510Z · LW · GW

Another thing to note that your Swedish & Finnish and English vocabularies might not have a proper inclusion relation. My technical & academic vocabulary is larger in English than in my native language, but for household appliances/vegetables/flowers it's the other way around. 

Comment by Self-Embedded Agent on Never Go Full Kelly · 2021-02-26T18:59:21.258Z · LW · GW

I initially gave this post a double upvote after a skim-read. The topic is of clear interest, and these ideas are new to me. Unfortunately, after taking a closer look I found the details difficult to follow. I decided to look at the linked paper. It seems you are trying to summarize this paper, perhaps this could be stated clearly at the top of the post.

Comment by Self-Embedded Agent on Weird Things About Money · 2021-02-24T14:49:51.739Z · LW · GW

Possibly relevant in the context of Kelly betting/ maximazing log wealth.

Comment by Self-Embedded Agent on Timeline of AI safety · 2021-02-09T08:43:12.667Z · LW · GW

Ah! Excuse me for my drive-by comment, I should have clicked the link.

Comment by Self-Embedded Agent on Timeline of AI safety · 2021-02-08T14:32:24.152Z · LW · GW

The effort is commendable. I am wondering why you started at 2013?

Debatably it is the things that happened prior to 2013 that is especially of interest.

I am thinking of early speculations by Turing, Von Neumann and Good continuing on to the founding of SI/MIRI some twenty years ago and much more in between I am less familiar with - but would like to know more about!

Comment by Self-Embedded Agent on I Was Not Almost Wrong But I Was Almost Right: Close-Call Counterfactuals and Bias · 2021-02-02T15:29:30.684Z · LW · GW

Datapoint: I skim-read the article today. I am interested in the overal thesis [need for closureness, counterfactual modelling etc]. I skipped the Dr. Zany story. 

Comment by Self-Embedded Agent on Against Sam Harris's personal claim of attentional agency · 2021-01-31T13:15:46.562Z · LW · GW

I see two views:

Sam Harris the Enlightened Guru, calm-voiced, leader,  unjudgemental, high status, above it all. 

Sam Harris a Flawed Mortal, full of judgements, self-denial, caught up in the Culture War.

I sense a conflict between these two views you might have of Sam Harris. Is there something to this?

Comment by Self-Embedded Agent on #2: Neurocryopreservation vs whole-body preservation · 2021-01-13T11:36:32.730Z · LW · GW

Uploading seems the most plausible route with very high probability to me. 'Reviving' a frozen body seems impossible: neurons in the brain will be severely damaged, there seems no plausible way to repair them without magic nanotech at which point uploading seems easier. 

Comment by Self-Embedded Agent on 2020's Prediction Thread · 2020-09-01T07:10:03.265Z · LW · GW

Good point..! I also recently came upon

I thought I remebered that some of the mechanisms were unknown.

Comment by Self-Embedded Agent on Was a PhD necessary to solve outstanding math problems? · 2020-07-10T20:32:42.874Z · LW · GW

It's really quite simple. Doing mathematical work at the highest levels requires both extraordinary talent and single-minded focus on mathematics. Mathematics is to some degree a younger man's game, yet modern mathematics requires knowing vast amounts of previous work to have any shot at solving a serious conjecture.

Doing a PhD is the most straightforward path to acquiring the knowledge to make a serious attack. Most fields will require upwards of a decade to build up enough technical experience to attempt to solve famous open problems.

Spending say 5 years acquiring money to do independent work doesn't seem like a good plan, when the people who have the talent required to solve one of these problems are almost guaranteed to obtain an academic position.

Comment by Self-Embedded Agent on [Reference request] Can Love be Explained? · 2020-07-09T12:39:15.782Z · LW · GW

This is what I was looking for. Thank you, Gwern!

The heart has its reasons that reason knows not of
Comment by Self-Embedded Agent on [Reference request] Can Love be Explained? · 2020-07-07T14:33:59.603Z · LW · GW

I mean romantic attraction.

Comment by Self-Embedded Agent on [Personal Experiment] One Year without Junk Media: Six-Month Update · 2020-06-21T16:32:18.187Z · LW · GW

One problem I have in keeping this kind of regime is that I often find myself going around my blocking software, e.g. through using private browsing. Do you have a recommendation for something that works more generally?

Have you tried software like RescueTime? I have used it in the past and was happy about it (it is a more comprehensive system than just plain blockers; also tracking what you are wasting time on; allowing differeent levels of usefulness etc)

Comment by Self-Embedded Agent on [Personal Experiment] One Year without Junk Media: Six-Month Update · 2020-06-21T08:25:07.111Z · LW · GW

Your experiment sounds very interesting. I would like to follow in your footsteps.

Could you detail for a digitally illiterate how you did this?

How long should I keep it up before I will feel results in your estimation?

What site plugin did you use?

Comment by Self-Embedded Agent on What is a decision theory as a mathematical object? · 2020-05-25T17:34:54.628Z · LW · GW

Fundamentally, finding a good mathematical definition of decision theory that encompasses all the phenomena people care about is a big open problem.

Comment by Self-Embedded Agent on [deleted post] 2020-04-20T19:42:46.998Z

Is this a joke? Genuine question. :) It seems English might not be your native language.

Comment by Self-Embedded Agent on How credible is the theory that COVID19 escaped from a Wuhan Lab? · 2020-04-04T06:23:34.644Z · LW · GW

Why is it an infohazard?

Comment by Self-Embedded Agent on Resource for the mappings between areas of math and their applications? · 2020-03-30T09:08:28.820Z · LW · GW

I will go against the advice that you were offered. Especially early on I think trying to understand applications can be a bit of a trap. Either the application is so simple it can be explained without the math[e.g. twisting a factory band into a Mobius strip to make the band wear on both sides, square-cube law, logistic curves in epidemics] or the details are actually quite complicated, which may obscure one's understanding of what is the actual generalizable math concept and what is specific to this problem.

The prototypical application of calculus is Newton's work on astronomy & mechanics. This is a typical case of the latter.

That said, I suppose you've heard of

[1] 3Blue1Brown

Comment by Self-Embedded Agent on March Coronavirus Open Thread · 2020-03-20T12:36:02.722Z · LW · GW

Just like to chime in to say that this (=' flattening the curve/ herd immunity') fundamentally doesn't work, and you don't need to have a PhD in epidemiology from Imperial College to understand this [but you might need a PhD in epidemiology to misunderstand it], just basic arithmetic and common sense.

Suppose 50% of the UK (33 million people) get the virus of which 5% (~ 1.8 million people) will need serious hospitalization [conservative estimate]. The current capacity of ICU beds in the UK is something on the order of 2000 beds , depending on occupancy rates, ability to scale up et cetera. Let's be extremely optimistic and somehow the UK is able to quintuple this capacity [as far as I can tell this is unlikely]. When somebody is sick they might need care for 2 weeks. The annual hospital capacity is: 25 weeks * 10.000 beds= 250k. At the moment the capacity is nowhere that (perhaps 50-100k).

You can see that 1.8 million is far larger than 100k or even 250 k. Even wildly optimistic estimates will not yield anything realistic. This assumes that the government is somehow able to control the infection spreading over a year; instead of two months. There is no reason to think they can do this without extreme (partial) lockdown measures. Controlling the R0 is extremely hard. All the mild measures seem to help only a tiny little bit. If the R0 is only a bit over 1, we still have exponential growth; and you have merely pushed timelines back a few months.

Can we perhaps expose young people but lock up older people for one-two years [when the vaccin might arrive]? I find this is extremely unlikely; you need only a couple people to flout the rules to wipe out an entire nursing home.

Is it worth it to (partially) lock down the entire country for a year to save maybe a hundred thousand old people?
There are only two real possible approaches:

1. Let the Boomers die. If we're lucky the death rate is ~0.7 percent. When (not if) hospitals overflow this will easily triple. Without medical care, once you go critical you simply can't breath [though I heard something to the effect that most/many deaths are due to cardiac arrest]. Simple as that. With a massive host population the virus will mutate and we might have the same problem every year [<- this very real possibility is perhaps the most important to think about].

2. total lock down -> squash the curve, followup with massive testing Gangnam style and extensive contact tracing [also: Fast-track all possible vaccins/treatments and fire Chief Medical Officer]. This seems to work so far in all East-Asian countries. Why the people with actual experience and succes in this matter get systematically ignored in these discussion will be a question for historians.

I know my preferred approach. There is no linear response to an exponential tide.


Comment by Self-Embedded Agent on The recent NeurIPS call for papers requires authors to include a statement about the potential broader impact of their work · 2020-02-24T10:30:35.367Z · LW · GW

Goodhart's Law poised to strike.

Comment by Self-Embedded Agent on 2020's Prediction Thread · 2020-01-14T04:53:04.200Z · LW · GW

Usual disclaimers apply: probabilities are not exact betting odds, I try to give quantitive assessments wherever I can but many predictions are too vague to quantify etc. If I am still alive in 2030 I will try to give my subjective assessment to what degree I agree with the predictions.

1. a China will become #1 economy in the coming decade, but will experience continued economic slowdown.

2. Taiwan put under siege by China (i) economically 80% (ii) militarily 50%
conditional on (ii) the US will back down 90%
conditional on the US not backing down this will lead to a nuclear exchange between US and China 10%

3. A missile/drone strike on an aircraft Carrier, crippling the ship 4%

4. India will commit what will later be regarded by a significant fraction of Western observers as genocide in Kashmir 10 % (low confidence)

5. Economic slowdown in the West (GDP/capita growth for 2020-2030 lower than 2010-2020) 70% - a plausible story is an aging population
6. Nuclear strike anywhere in the world 5%
7. a Continuing secularisation in the Middle-East 75%
b continuing secularisation in the West 85%
8.a Subsaharan Africa continues making little gdp /capita gains 70%.
b Population will continue to boom. Fertility will fall modestly 70%
c Immigration streams to the West from 3rd World countries, particularly Africa will increase 70%

9. Smoking/drinking will continue to fall worldwide, party drugs (MDMA etc) will become much more widespread 70% & 70%
10. No economic Recession comparable with the 2008 Great Recession 85%

Politics & Culture Wars
1. Revolution in Energy will increasingly make present discussion around climate change irrelevant

2. a Trump does not win reelection 70%
b conditional on Biden winning the primary, no crazy things (i.e. somebody dieing) Biden wins 85 %

3.a Conditional on it happening, Brexit impact on UK/England GDP per capita long-term will be negligible (< 2% of GDP counterfactually)

b conditional on (a), there will still be a significant fraction of people who honestly think that Brexit was a disaster. Most people
4. Higher education bubble in US will subside (echoing Vaniver, hard to make it more concrete, more know it when you see it)

5.a Traditional media will continue to fall into irrelevance, concentrate in a few major firms. Later into the decade I expect the decline to halt, traditional media taking a specialized role for particularly high authority/ high importance news. I model this as similar to Radio still having specialized uses.
b (some) Youtube& Instagram stars will attain widespread fame and recognition in a much broader context as the Millenial generation moves towards middle age. But I do not expect them to displace current celebrity elite based on tv&movie stars, musicians, sports heroes, etc.
c followup on b), the fragmentation of assigned status will continue in the West [but not necessarily globally, as globalization homogenizes cultures].
6. What will happen with the Culture Wars? This is very hard to predict. Will the apparent trend of increased polarisation continue, or will the opposite occur?
- Straight Woke memes will become less cool with the incoming generation of young people, mostly because whatever your parents do isn't cool. But the ideas of social justice will continue their expanse, particularly in the Middle east, South America, South-east Asia & India: social justice memes are fundamentally fit in a world of increasing wealth & communication. Another fundamental force is the continuing expanse of English: new media like youtube, continuing globalization etc have given renewed impetus all over the world to learning English. Of course these movements generate their own antitheses; it seems like historically many conflicts arise from a modernizing center versus a lagging periphery.
China & East Europe seem more like a wild card here: eastern Europe seems to have strong explicitly anti- Social justice political movements while China is a surveillance state with very strong hard-Nationalistic rightwing sentiments under the population.
- Trans-rights increasingly accepted
- Alt-Right ideas increasingly normalized (Likely) - It seems that moderate right wing parties are losing hearts and minds against more radical alternatives.
- Identity politics/ voting along identity continues in strength (Likely)
- increase popularity national unity/ decreased polarisation candidates in US. After Trump I expect a backlash towards more moderate, unifying candidates.
- Right-wing populist parties in Western Europe continue to grow, will remain mostly out of governments.
but at least one right-wing populist party will be in government in a Western European country that is not Italy

1.a More babies born with edits for medical reasons 85%
b more babies born with edits for non-medical reasons 70%
c babies born with explicitly IQ edits 60%
d CRISPR or CRISPR like techniques cure >5 genetic disease 70%
2. a Discovery cure for (1) large fraction of obesity (2) large fraction of heart disease (3) Male Pattern baldness (4) Dandruff 10/10/20/20

b Conditional on a) this cure is available on the market 30%
c conditional on a) this cure works by defeating some sort of germ/biological agent 20%/30%/80%/80%
3. Note: by mini-drone I mean a small drone that one would buy in the store, not the unmanned aerial vehicles and its siblings that the US military employs.
a) mini-Drone attacck will kill major Political Figure 85%
b) mini-Drone/drone bombs will be used in major terrorist attack killing many civilians 85%
c) conditional on a or b, massive fear around drone warfare
d) UAVs will become more and more dominant.
e) Something I am not sure about: will thee US military start phasing out fighter aircraft for automated vehicles or is this perhaps one of those jobs that is very hard to automate?
4. Crazy person kills AI researcher because of vaguely AI risk type concerns 80%
5. Top ranked Starcraft Broodwar player trounced by AI by 2025 (70%) by 2030 (85%)
6.a SpaceX sends that Japanese Billionaire around the moon 55%,
b conditional on a launch, it goes wrong 10%
c SpaceX will land unmanned probe on mars by 2030 85%
d SpaceX will land a human on Mars by 2030 60%
e SpaceX will become wildly successful and highly valuable with Starlink technology 60%

f Tesla will be the largest automotive company by market cap at least once 70% [this is the only prediction that I can actually bet on, but I do not own Tesla stocks... you decide what that means]
7. Self-driving cars will finally become a practical reality -?? very uncertain, I have very little knowledge about this, I see convincing arguments for both sides. Certainly it seems that self-driving cars in some capacity - perhaps trucks driving pre-determined routes- will become a reality, but I remain uncertain. I predict I will update strongly one way or another on learning more about this subject.
8. Exoplanet that is an almost copy of earth discovered (i.e. habitable zone, water, same size etc)- Likely too vague to give quantitative estimate.
9.a) An AI will compose a hit song without significant assistance. (80%)
b) AI-assistance in arts/creative professions such as writing (fiction, non-fiction, poetry, journalism), music, as well as in daily life will be widespread.

10. DeepFakes will be huge.
a) porn with DeepFakes will start seriously competing with real life performers
b) remix art with deepfakes; i.e. reshooting the (dissapointing) ending of popular tv dramas, completely redoing old films, etc etc.
c) value of videoevidence will be much less persuasive. This will have massive implications for court proceedings.

11. AI Autumn - the spectular developments we have seen recently with deep learning and related methods are unlikely to be equalled in the coming decades [just a regression to the mean argument here + there was somewhat of a compute overhang that is unlikely to be repeated] but the space of things to try seems quite large + the field is well-funded and attracting a lot of talent.

12. No Superintelligence by 2030; pretty clear. 95% Mostly a prior on complexity, reaction against hype etc.
13. a Satoshi Nakomoto (briefly) richest person in history 30%
14. >80% of cars on the road in US will be electric 80%
15. Working anti-aging intervention for mice 20% , working anti-aging for humans 10%, on the market 5%.
16.a VR entertainment will be the majority of the game market in terms of revenue - 70%
b VR entertainment will be a decided minority in terms of time spent playing. - 70%.
c VR porn will become a thing. Easy prediction here.
17. Military-grade lasers will evolve from mostly testing prototypes to usable weapons - somewhat likely

1. In general, Rationalist memes will become much more mainstream

a) Existential risk, in particular AI x-risk will be regularly mentioned by mainstream media- Likely
b) AI safety more generally will be a topic that commands at least 10% of the attention that global warming does - Likely

c) increased acceptance of cryonics & assisted suicide -Likely
d) HPMOR and the associated rationalist memeplex will be atleast well-known enough to need little explanation in high-brow circles.

22.a) AI Safety will be a respected academic discipline 90%

b) MIRI-style research will be a decided minority within this discipline 80%

c) MIRI will not continue their nondisclosure policy in the present form 70%

23. Effective Altruism mainstreamed significantly - 85%
Miscellaneous & Personal 1. I will become a parent 50 %
2. I change my specialization significantly 40% [It shocks me how high this is if I apply the outside view... generally applying outside view to oneself is distressing...]
3. I will die 1.28 %
4.a) Proof-assistants will be part of undergraduate math curriculum - not sure about this one; it is being done right now in an undergraduate program in London, and it seems to make sense. The undergrad math curriculum is already pretty full as is. Somewhat likely.
b) (Homotopy) Type theory will be offered as a regular module in good European Mathematics/Compsci Bachelor or Master's programs Likely

5. Programming will be offered as a serious stand-alone course at my high school. 65%
6. Significant developments in understanding agents based on Compositionality- somewhat likely

Comment by Self-Embedded Agent on Is cardio enough for longevity benefits of exercise? · 2020-01-11T00:48:42.285Z · LW · GW

People report feeling better because of all kinds of reasons. This line of reasoning seems unlikely to convince a skeptic.

Let me throw out an alternate hypothesis, which is a little extreme but ought to be considered: doing sports is a form of health signalling. People signal their health and conscientiousness by sports. Let's call it the Health Causes Sports theory, as opposed to the Sports Causes Health theory. Notice that this neatly explains people feeling good after sports: they succesfully send a difficult-to-fake signal.

A problem with the Sport Causes Health theory is the weak evolutionary story. Why would expending energy and risking bodily harm for no concrete payoff be good for one's health? Many sports carry moderate health risks. Wouldn't it make more sense to not have health tied to activity?

The strongest argument is the Greasing the Gears - theory: by streneous activity people grease their bodily machinery. Too much is bad, but so is too little, and modern humans do too little of it. If this theory is right, it certainly doesn't seem to be any specific muscle [ as the precise sport doesn't seem to matter] but a generalized cardiovascular capacity.

It seems true that modern peoples are much more sedentary than in the past. On the other hand, the amount of activity that people display likely varies wildly over societies& time & place. The 'optimal level of activity' likely also varies with genetic background. If there really was a need to grease physical machinery one would think that the body would try to automatically modulate this, in the same way it has a thermostat for many important physical quantities.

The best argument for the Sport Causes Health theory is probably rats: exercising rats seems to increase their lifespan. There is a plausible story that this has something to do with mitochrondial capacity [I recommend de Grey's book Ending Aging and Nick Lane's 'Mitochrondia' for some possible theories]. Yet lots of things increase lifespan of rats, most famously calorie restriction. Here too, there is a plausible sounding mitochrondial story, as well as other stories. Yet the consensus is that the lifespan increase for humans using calorie restriction is relatively small, if not wholly insignificant [There was a recent post on SSC about this; my takeaway was that even if the effect exists it absolutely wasn't worth it]. Generally, only a tiny fraction of experiments with mice/rats replicate in humans [I've heard 10% being bandied around].

Most supercentenarians don't seem to have engaged in heavy sports, while top sportsmen & women do not exhibit exceptional lifespan increases over the general population. More general, the effect of iq on lifespan is much larger, suggesting underlying biological/genetic reasons dominate for lifespan.

Does this explain all health variance with regard to sports? I would say that's pushing it, yet I certainly would like to see this alternate hypothesis considered. I think there are some good arguments to be made for both theories (Sport Causes Health, Health Causes Sports). All in all, my best guess is that very moderate amount of activity is likely beneficial for health, but sports in excess of that is mostly health-signalling.

See also Hanson about signalling in team sports/competitions:

Comment by Self-Embedded Agent on Is cardio enough for longevity benefits of exercise? · 2020-01-04T18:56:17.146Z · LW · GW

Twin studies make my heart sing, so if you have a link I would be grateful!

Circulation and stress regulation are rather vague. There is more blood flowing, so suddenly your neurons are long-term better working? How is this consistent with (i) the blood-brain barrier and (ii) the general observation that cognition cannot be reliably be improved by interventions? Also, what is the evolutionary story?

The sleep quality seems anecdotally somewhat real. If you follow the links above otoh much of the research seems rather iffy.

In general, references would be appreciated.

Comment by Self-Embedded Agent on Is cardio enough for longevity benefits of exercise? · 2020-01-03T21:45:19.193Z · LW · GW

I am incredibly skeptical of the supposed health benefits of exercise in general.

My alternate hypothesis is that healthy people exercise more because they are healthier.

Of course, there are some interventional studies but I am increasingly skeptical of the general competence of researchers in nutrition/health; my null hypothesis is simply that these studies are confounded/done badly/p-hacked/ etc etc.

One of the supposed benefits quoted in the link above is cognitive benefits. Although I can imagine some limited health benefits from exercise I am completely incredulous of cognitive benefits. What is a plausible mechanism of action where training your muscles has any effect on your brain, that is moreover evolutionarily sensible?

If anybody could provide strong unambigous evidence for these claims I would love to hear them.

Comment by Self-Embedded Agent on Market Rate Food Is Luxury Food · 2019-11-24T03:29:37.588Z · LW · GW