This is a literary way of saying I am not as interesting as my writings, and in some respect, it should not matter who I am or what I have done because argument screens off authority.↩︎

When I say “research assistant”, I mean it in the older sense of someone who does detail work for another person’s original research—so I spent a lot of time reading up on specific areas and making notes about stuff my boss needs, and only occasionally do independent work. Not all my work can be made public, but some of it is. A partial list in rough chronological order:

• “Intelligence Explosion: Evidence and Import”, Muehlhauser & Salamon 2012_12ya?

  One of my first projects was working on this paper—digging up citations for various claims and giving general feedback; most of my changes or comments are private when we switched to working through Google Docs, but some of my initial comments were public.

• finding predictions/estimates for whole-genome sequencing costs

• “The Personality of (great/creative) Scientists: Open and Conscientious”

  Summarizing some of the existing research; besides being intrinsically interesting, this is also relevant to computer intelligence—if achievement is just a reflection of IQ, then any computer intelligence will improve on humans only to the extent it has greater raw intelligence, but if achievement is limited past 130s IQ, then the picture becomes more complex. (Computer intelligence may be far more more powerful than predicted for a given intelligence level because there’s no apparent reason you couldn’t easily make its ‘personality’ open-minded—it not really being clear that truly random exploration and computing power don’t yield creativity in some fields already—and a computer can be tireless and self-disciplined, which takes care of Conscientiousness. On the other hand, Openness and Conscientiousness may correspond to subtle high-level dynamics which can’t be easily designed into an AI and can’t be produced by throwing computing power at the problem, in which case the AI will underperform whatever measure of intelligence we ultimately use, like AIXI-style IQ tests.)

• Remote monitoring

  Some quick notes on remote monitoring software MIRI could use to monitor freelancers such as myself. (It’s a good thing, really—trust but verify.)

• Some info on the big Beijing genomics project, and estimates about personality and IQ, and past Chinese abuse of genetic information

• Are Sunk Costs Fallacies?

  A detailed examination of one cognitive bias asking whether it has been misinterpreted as a problem for individuals.

• “Against Utilitarianism: Sobel’s attack on judging lives’ goodness”

  Philosophy essay summarizing and attempting to analyze a recent class of objections to utilitarianism (that one cannot judge utilities of lives, and hence the utilitarian ethical project is impossible).

• Neuroscience additions to Wikipedia; I had a hard time finding references for a MIRI paper, and donated my results to Wikipedia so at least no one else has to suffer so much:

  • Exactly how big are chimpanzee & human brains, anyway?

  • Primate and human brain size correlates with IQ (No, really. The correlation isn’t terribly big, though.)

  • Rat IQ (mixed evidence, thankfully)

• polyamory review

  Not really enough data to do anything but point out problems for reliable assessment. My subjective impression is that polyamory is complex enough that it may not be a good idea even for polyamorists.

• “Cashing Out Cognitive Biases as Behavior”

  Spin off from “Are Sunk Costs Fallacies?”—2 studies examining consequences in personal life from vulnerability to cognitive biases in questionnaires

• “Experiment: a good researcher is hard to find”

  A simple email experiment on making applicants feel that their admission test is not a waste of time: they seem to do better work.

• Geopolitical intelligence failures—notes from some reading in the history of intelligence failures

• “Notes on the Psychology of Power”

• “Notes on Psychopathy”

• “Notes on Autonomous Cars”

• LSD & Modafinil and schizophrenia

The following is a list of my submissions to LW I regard as substantive or particularly good, excluding content which can be found on Gwern.net, in chronological order with interesting ones highlighted:

• “Poker Playing”

• “Lifelogging: the recording device”

• “AIXI-style IQ tests”

• “Is Kiryas Joel an Unhappy Place?”

• “SL4 META: list closure 2 month followup”

• Life extension:

  • “Living Forever is Hard, or, The Gompertz Curve”

  • “Living Forever is Hard, part 2: Adult Longevity”

  • “Prediction is hard, especially of medicine”

  • “Living Forever is Hard, part 3: the state of life extension research”

• “(Philosophical) Disagreements are not Rational”

• “A good volunteer is hard to find”

  followup: “Experiment: a good researcher is hard to find”

• “On the unpopularity of cryonics: life sucks, but at least then you die”

• “How to be Deader than Dead”

• “Case Study: Reading Edge’s financial filings”

• “Atheism & autism spectrum”

• “Stanislav Petrov Day”

• “Not By Empathy Alone”

• “On the Openness personality trait & ‘rationality’”

• “Social status & testosterone”

• “Amanda Knox: post mortem”

• “Cryonics costs: given estimates are low”

• JET paper reviews:

  • “‘Ray Kurzweil and Uploading: Just Say No!’, Nick Agar”

  • “‘Vulnerable Cyborgs: Learning to Live with our Dragons’, Mark Coeckelbergh”

  • “‘Misbehaving Machines: The Emulated Brains of Transhumanist Dreams’, Corry Shores”

  • “‘Personal Identity and Uploading’, by Mark Walker”

• “Does Hyperbolic Discounting Really Exist?”

• “Cryonics is Far, Cord-blood is Near”

• “Inverse p-zombies: the other direction in the Hard Problem of Consciousness”

• “The Personality of (great/creative) Scientists: Open and Conscientious”

• “The Conditional Fallacy in Contemporary Philosophy”

• “Cashing Out Cognitive Biases as Behavior”

• “On the etiology of religious belief”

• “To Learn Critical Thinking, Study Critical Thinking” -(general philosophy studies are not as efficient in improving critical thinking as critical thinking courses)

• “Against NHST” -(the statistical controversies over the p-value paradigm)

• “Notes on Brainwashing & ‘Cults’”

Of course, I don’t agree with every MIRI or LW position. The intellectual homogeneity has been much over-estimated by outsiders who have not bothered to look at the annual surveys, I think. Here are some major points for me:

1. MWI: I think that LWers who were persuaded by Eliezer’s MWI writings are wrong to do so, as they are unfamiliar with even the rudiments of any alternatives interpretations and cannot judge in the matter; how many LWers have ever seriously looked at all the competing theories, or could even name many alternatives? (“Collapse, MWI, uh…”), much less could discuss why they dislike pilot waves or whatever. Lacking any real understanding, they ought to simply adopt the expert consensus, where MWI seems to have a plurality or bare majority of adherents (with the weak confidence that implies).

2. Heuristics and cognitive biases: I am not much convinced that knowledge of heuristics & biases help in ordinary life. Feedback & learning are powerful tools in eliminating error, calibrating predictions, and justify committing what may look like the sunk cost fallacy; and feedback is what one gets in ordinary life.

   Per Moravec’s paradox, where our knowledge of heuristics & biases will pay off most is in what Hanson would call “Far” scenarios: evolutionary novel situations with few precedents and only costly or non-existent feedback. (For example, the question of whether artificial intelligence will be developed by 2040: it will only happen or not once, there are few comparable events, the consequences may be dramatic, and our ordinary lives offer no useful insights.) As it happens, this describes much of futurism & forecasting but we cannot justify our futurism by claiming its techniques are incredibly valuable in ordinary life!

3. Cryonics girl: The donations appall me, for reasons I lay out at length there—they are a complete abandonment of core ideas like utilitarianism & optimal philanthropy.

4. Alicorn’s “Living Luminously” paradigm struck me as dubious, not backed by even token research, and likely idiosyncratic to her; I thought her Luminosity e-novel was merely OK despite the endless discussions on LW (rivaling those for Methods of Rationality itself) and that her followup, Radiance, was just terrible. Nevertheless, her novel career seems to continue.

There is a moderately funny story about how Gerard came to write it, based on my musical incompetence.↩︎

That is, summing up the (surviving) edits of my various accounts over the years: User:Gwern, User:Marudubshinki, & User:Rhwawn↩︎

Compare the CSE results with the Google Results for the anime Wings of Honnêamise. Which is more useful for an editor? For more details, see my release announcement.↩︎

A trick I discovered when visiting FHI in 2015—I had used widescreen laptops for so long I had forgotten how nice portrait-orientation was for reading.↩︎

I had a Kinesis Advantage keyboard, but struggled with the keymapping & large physical size making it difficult to find a comfortable desk & chair height which left my arms high enough to use the thick Kinesis. In July 2020, I switched to the split ergonomic mechanical keyboard Ergodox EZ ($325), but that wound up having similar issues.

At this point, I began getting frustrated with the time & money I was spending dabbling in exotic keyboards—the biggest problem with the generic keyboard was the switches, and that the ten-key island took up a lot of space & made it hard to reach for the trackball. So in September 2022, I looked through a keyboard search site until I found a cheap $50 thin ten-key-less keyboard with good low-travel mechanical switches (to help forestall RSI from heavy keypresses) from G.Skill, and called it a day. It is thin enough to fit my posture, didn’t require painful relearning of decades of muscle memory, and has been satisfactory.↩︎

My best guess is that my problem initially was that I seriously underestimated how much pressure it takes to insert a Threadripper CPU into its socket—it required a truly terrifying amount of force and I only got it right after triple-checking online tutorials & videos & discussions—and that was why the first motherboard never worked at all, and the second one was killed by static electricity or a short.↩︎

See also “Actively Open-Minded Thinking Scale”, “Clarity Scale”, “Engagement with Beauty”, & “a measure of what types of stories you enjoy”.↩︎

See also “Zimbardo Time Perspective Inventory”. Brent W. Roberts criticizes these two inventories when used to measure Conscientiousness.↩︎

See also “Relational Mobility scale”, “Empathizing and Systemizing scales” & “Rational vs Experiential Inventory”.↩︎

See also “Self-Report Psychopathy Scale”.↩︎

See also “Experience in Purchasing Behavior Scale” & “Kentucky Inventory of Mindfulness Skills”.↩︎

For further reading on overconfidence, see all LW articles so tagged. I once read in a book of a study in which subjects were asked to generate ideas for, IIRC, putting out a fire, and to stop only when they were convinced they had thought up all good ones, and usually stopping when they had thought up only a third; but I have been unable to refind it and would appreciate knowing details if this description rings any bells for a reader.↩︎

Chapter 17, Tao Teh Ching↩︎

For example, my clean-up and extension of the browse-url module was completely rewritten by RMS; so I can hardly take credit there.↩︎

Henceforth, this implies I have a commit-bit (or equivalent) for that project.↩︎

Henceforth, ‘cleanup’ should be taken as referring to extensive miscellaneous changes which include (in no particular order):

• fixing GHC’s -Wall or hlint warnings

• replacing OPTION pragmas with LANGUAGE pragmas

• tracking down licensing information

• switching from Haskell98 imports to the standard hierarchical module imports

  1. eg. import Char → import Data.Char; nontrivial in some cases where Haskell98 modules were dispersed over multiple base modules

• reorganizing the file tree

• improving the Cabalization

• whitespace formatting, and so on.

Henceforth, this typically implies that I uploaded it to Hackage as well↩︎

Henceforth, this implies that I made whatever changes necessary to get it compiling on GHC 6.8.x and 6.10.x↩︎