Measure's Shortform 2020-11-27T05:22:34.678Z


Comment by Measure on “Who’s In Charge? Free Will and the Science of the Brain” · 2021-09-16T20:28:26.610Z · LW · GW

FWIW, under the Many Worlds Interpretation quantum physics is just as deterministic as classical physics.

Comment by Measure on How to deal with probabilities in the presence of clones? · 2021-09-16T20:21:02.984Z · LW · GW

Immediately upon finding myself in the room, I could be the single copy in scenario B or any of the 8 copies in scenario A. I think I would have to assign 8:1 odds for A:B. Either way, I have a 7/8 chance of dying after a minute. After 1 minute, there is 1 clone in scenario A and 1/8 of a clone in scenario B, so I'm still at 8:1 odds for A:B.

Comment by Measure on Optimizing Multiple Imperfect Filters · 2021-09-16T01:20:26.681Z · LW · GW

Make the cutoff very high, and every widget fails the test - no false positives, but a very high false negative rate.

This is reversed. (The following sentence is correct.)

even though they would have bought

A couple of errors here.

Comment by Measure on DanielFilan's Shortform Feed · 2021-09-15T15:01:48.925Z · LW · GW

My 5-second take is basically what Gunnar_Zarncke already said. If you're finding difficulty writing something clearly, it might mean you don't understand it as clearly as you think. Maybe you understand 90%, and you gloss over the unclear 10%. Writing it out (or trying to fully explain it to someone) forces you to work through that 10%.

Comment by Measure on MikkW's Shortform · 2021-09-12T20:45:47.736Z · LW · GW

I think you mean a queue rather than a stack.

Comment by Measure on MikkW's Shortform · 2021-09-12T03:41:37.463Z · LW · GW

I hadn't previously heard of this method of combining candidate-within-party and party-level preferences into a single vote. Seems like a nice, simple system.

Comment by Measure on Measurement, Optimization, and Take-off Speed · 2021-09-11T14:53:23.283Z · LW · GW

Measurement Is Great

I approve of this.

Comment by Measure on Prefer the British Style of Quotation Mark Punctuation over the American · 2021-09-11T14:44:38.273Z · LW · GW

An argument could even be made for using a second period to close the outer sentence after resolving the inner sentence and the quote.

Dr Johnson kicked a large rock and said, as his foot rebounded, "I refute it thus.".

This looks weird to me because it's not what I'm used to, but it makes logical sense.

Comment by Measure on Bayeswatch 9: Zombies · 2021-09-11T08:04:12.156Z · LW · GW

look enough -> long enough

Comment by Measure on The Blackwell order as a formalization of knowledge · 2021-09-10T16:01:47.527Z · LW · GW

Blackwell’s theorem says that the conditions under which κ1 can be said to be more generally useful than κ2 are precisely the situations where κ1 is a post-garbling of κ2.

Is this reversed?

Comment by Measure on D&D.Sci Pathfinder: Return of the Gray Swan Evaluation & Ruleset · 2021-09-10T00:52:22.555Z · LW · GW

I had a lot of fun with this one. There were a lot of things I missed, and I could have spent a lot longer looking for patterns if I'd had more spare time. The 2d + time meant a lot of room to search, but this was partially offset by sensible real-world structure (e.g. northern icebergs, reefs near land, etc.) I agree with Yonge that this level of complexity was nice for a change but would be a bit excessive as the norm. I used a combination of Excel and Python and didn't have as much trouble with the data format as others seem to have had.

As far as my strategy, I started with routes that avoided previously dangerous hexes (my eastern route was a bit long, but I was hoping that the two ships sailing together could support each other or at least increase the joint probability by correlating encounters). I missed the migrating merfolk, so I avoided the shorter northern route.

I didn't spend as much time selecting ships. I noticed that Dhows had better survival rates but missed that they were assigned easier voyages. I didn't notice the captain effect on reef encounter rate and chose Dhows to avoid the high per-encounter damage given that I had a lot of unavoidable reefs on my routes. I chose ships that had a long history of successful voyages and did manage to get two seamanship-3 captains.

Comment by Measure on A Primer on the Symmetry Theory of Valence · 2021-09-07T14:02:48.765Z · LW · GW

Waves and symmetries don't carry many bits of information. If you think valence and suffering are fundamentally few-dimensional, maybe that doesn't bother you; but I think it's at least possible for people know whether they're suffering from arm pain or finger pain or air-hunger or guilt or whatever.

"What is the problem?" should have a pretty high information content, but there might be a separate "how bad is it?" question that constitutes the actual unpleasant part of the experience, which wouldn't have to be much more than a 1d scalar.

Comment by Measure on Obstacles to gradient hacking · 2021-09-06T11:54:46.212Z · LW · GW

One particular formalism that might work is continuous everywhere and differentiable everywhere except a set of countable size.

I'm confused by this part. If the function is stored numerically in network weights rather than symbolically, how do you verify that it has these properties?

Comment by Measure on Acausal Trade and the Ultimatum Game · 2021-09-05T12:06:29.290Z · LW · GW

Against a real partner — not a copy of yourself — it's better to leave some wiggle room rather than simply rejecting offers less than 50% in case your partner has a slightly different notion of fairness from yours. For example, you could reject increasingly lower offers with increasing probability such that their expected utility is maximized at 50%.

Comment by Measure on [deleted post] 2021-09-04T17:03:38.089Z

The short answer is that if you have a set of mutually exclusive hypotheses, such as a proposition and its negation, then your credences for them should sum to unity. Conditional on MTBI's assumptions, a 55% credence for T-type does indeed imply a 45% credence for F-type. If I think people will misunderstand me due to connotations, then I need to be more clear when expressing myself, but that doesn't change my internal beliefs.

Comment by Measure on D&D.Sci Pathfinder: Return of the Gray Swan · 2021-09-03T20:18:22.527Z · LW · GW

Route Selections:

(Red Duck) Q6-P7-Q8-Q9-Q10-P11-O11-N11-N12-M13-M14-L14-K14-J14-I14-H13-H12-G12-F11-F10-E9-E8 (return same route)
(Scurvy Hind) Q6-P7-Q8-Q9-Q10-P11-O11-N11-N12-M13-L13 (return same route)

Other Notes:

Delay between voyages doesn't seem to affect damage.
Avoid the wizard on J8.
Avoid reefs when possible (adjacent to most land hexes except around ports).
Dhows take less damage per hex (most of this might be from dodging reefs).
Galleons attract Sharks/Pirates/Harpies/Merfolk/Dragons.

Comment by Measure on D&D.Sci Pathfinder: Return of the Gray Swan · 2021-09-03T13:54:18.837Z · LW · GW

If both ships have the same initial moves, will they get the same encounters in those hexes?

Comment by Measure on Lakshmi's Magic Rope: An Intuitive Explanation of Ramanujan Primes · 2021-09-03T00:05:50.789Z · LW · GW

You say there is a lower bound on the number of primes on the line, but in fact there are many. Any number not greater than the full number of primes is a valid lower bound. What you are looking for is the greatest lower bound that is also a lower bound for all higher/longer magic lines.

Pacing/Presentation Suggestion:
I was very confused when you were talking about counting primes before you mentioned the rule for the line length. If you'd started by explaining how long the line is (which is kind of arbitrary), then the discussion of the lower bound for primes would have been easier to follow.

Comment by Measure on D&D.Sci Pathfinder: Return of the Gray Swan · 2021-09-02T17:42:11.508Z · LW · GW

There are two ships in port captained by Angus MacDougal. Am I unable to use both?

Comment by Measure on A Layman’s Guide to Recreational Mathematics Videos · 2021-09-01T14:10:33.698Z · LW · GW

Zach Star has some great math videos as well.

Comment by Measure on JBlack's Shortform · 2021-08-28T13:05:27.894Z · LW · GW

Alternatively: The subject is show the result of the coin flip. A short time later, if the coin is heads, her memory is modified such that she remembers having seen tails.

Memory erasure can produce this sort of effect in a lot of situations. This is just a special case of that (I assume the false memory of a gentle waking also overwrites the true memory of the abrupt waking).

Comment by Measure on The Validity of Self-Locating Probabilities (Pt. 2) · 2021-08-27T17:32:50.589Z · LW · GW

That's the average of what all the clones will have at the end. The original should have about 250.

Comment by Measure on For me the Christianity deal-breaker was meekness · 2021-08-26T17:19:11.673Z · LW · GW

A lot of the reason for meekness, "turn the other cheek", etc. is based on the promise of reward in the afterlife. Without that incentive, I agree that valuing meekness makes less sense.

Also, there's a difference between not standing up for yourself and not standing up for others. In my understanding only the former is considered a virtue.

Comment by Measure on The Validity of Self-Locating Probabilities (Pt. 2) · 2021-08-25T19:49:05.602Z · LW · GW

I don't see why that would change anything. (This is all assuming my utility in gold bars is linear. Otherwise I'd have to worry about being flat broke and unable to feed myself if I turn out to be the clone.)

Comment by Measure on The Validity of Self-Locating Probabilities (Pt. 2) · 2021-08-25T18:34:15.441Z · LW · GW

After waking up, I hold P(heads) to be 1/3, so the bet is negative EV for me personally.

Comment by Measure on 18 possible meanings of "I Like Red" · 2021-08-25T16:39:33.070Z · LW · GW

[1]: Here, mean means "attempt to convey"

It seems like you're not using this definition consistently. For example, in (3), the speaker doesn't care that any information at all is conveyed since the purpose is reinforcing their own belief. Several of the others appear to be beliefs that would motivate someone to say "I like red" rather than what they intend to convey to the listener (in (7) what is intended to be conveyed is "I'm committed to support you.")

Comment by Measure on The Validity of Self-Locating Probabilities (Pt. 2) · 2021-08-25T16:21:53.489Z · LW · GW

P(original) = 2/3
P(heads) = 1/3
P(heads|original) = 1/2
Value of bet is (1/3)*5 = 1.667 < 2, so don't take the bet.

The policy of not betting maximizes the total payoff to all clones (6 vs. 5). The policy of betting maximizes the average payoff-per-clone over worlds (2.5 vs 2). If I don't care at all about my (possible) future clones, I should pre-commit to taking the bet, since this maximizes the payoff of the original. However, after waking up, I have to consider the possibility (with its associated probability) that I am the clone, so that changes my answer.

Comment by Measure on The Validity of Self-Locating Probabilities · 2021-08-24T14:25:01.380Z · LW · GW

I would say POI applies here.

Comment by Measure on The Validity of Self-Locating Probabilities · 2021-08-23T19:14:49.253Z · LW · GW

It's not that I wouldn't care which one I am (they're identical), but there would be no way for me to differentiate the experiences.

Suppose the scientist told me before the procedure that he would with probability 2/3 waken the original in a blue room and the clone in a red room and with probability 1/3 reverse the colors. If I were to wake up in a blue room afterward, my credence that I'm the original would be 2/3.

Comment by Measure on The Validity of Self-Locating Probabilities · 2021-08-23T13:43:14.624Z · LW · GW

I'm using "probability" "credence" and "betting odds" mostly interchangeably to refer to my subjective state of knowledge. The 50% number comes from the symmetry of having two indistinguishable experiences being had by two people, one of which is me. Without any additional information to break that symmetry (such as learning that multiple copies were created or that the copying sometimes fails), I should assign equal credence to each possibility.

Comment by Measure on The Validity of Self-Locating Probabilities · 2021-08-21T03:44:25.918Z · LW · GW

The copy and the original are indistinguishable to me/us, but as long as there is still a fact of the matter of which is which — the scientist knows after all — I would say it still makes sense to talk about the probability that I am the original. It would be exactly the same as asking "What is the probability, given that he answers truthfully, that the scientist will say I am the original when I ask?". I will get a definite answer one way or the other, so I think it makes sense that I should be able to have a credence for each outcome beforehand (50% in this case).

Comment by Measure on Covid 8/19: Cracking the Booster · 2021-08-20T16:46:34.278Z · LW · GW

Fair enough. I wasn't playing standard at the time, so my information is second-hand.

Comment by Measure on Covid 8/19: Cracking the Booster · 2021-08-20T13:53:10.056Z · LW · GW

Oko, Thief of Crowns was poorly balanced for the standard meta he was released into and ended up being very strong/oppressive due to being hard to interact with. The meme is that you wanted to just [play normal Magic]/[live your life], but you ended up having to focus on how to deal with [Oko]/[Delta].

Comment by Measure on Absent-Minded Driver and Self-Locating Probabilities · 2021-08-18T19:31:15.036Z · LW · GW

Each of A,B,C,D are measurable, and so AB+CD is measurable as well, but since that combination isn't the expected value we want to be maximizing, I would certainly say it's not useful and at that point I wouldn't really care whether it's meaningful (if it does "mean" anything, it would be something convoluted and not really relevant to the decision).

Comment by Measure on Absent-Minded Driver and Self-Locating Probabilities · 2021-08-17T20:38:56.543Z · LW · GW

What I mean is that you can just calculate each of...

A: Fraction of decisions made at X
B: Average payoff of decisions made at X
C: Fraction of decisions made at Y
D: Average payoff of decisions made at Y
E: Average payoff of all runs (some of which have multiple decisions)

...and observe that AB + CD = E does not hold. The reason for this is that some payoffs are the result of multiple decisions.

The overall probability for Heads is not the simple combination of probability "today is Monday" times p(Heads|Monday) plus the probability "today is Tuesday" times p(Heads|Tuesday).

I'm not a double-halfer, so I don't have a problem denying this.

Comment by Measure on Framing Practicum: Dynamic Equilibrium · 2021-08-17T04:01:24.809Z · LW · GW

lower...leave faster

This is backward.

Comment by Measure on Absent-Minded Driver and Self-Locating Probabilities · 2021-08-16T19:46:02.079Z · LW · GW

When I actually run the experiment many times, I find the following when using p=4/9:

  1. The fraction of decisions made at X is indeed 1/(p+1), in this case 9/13.

  2. Of all the times a choice is made at X, the average resulting payoff is 96/81 (as expected).

  3. Of all the times a choice is made at Y, the average resulting payoff is 24/9 (as expected).

  4. The overall average payoff is not the simple combination of probability at X times payoff at X plus probability at Y times payoff at Y.

The discrepancy is cause by the driver-at-Y having to "share" a portion of the expected payoff with the previous driver-at-X.

Comment by Measure on D&D.Sci August 2021: The Oracle and the Monk · 2021-08-16T18:08:35.764Z · LW · GW

My analysis:

Solar and Lunar both have a ~27 day high-low cycle (in phase with each other). There is an anomaly in the Solar data staring around 7 days ago (the supernova probably) that briefly caused Solar to spike (replacing the scheduled low point in the cycle). Lunar was unaffected by the anomaly, and the next peak for both should line up with the ritual day. Using [Solar + Lunar] for the ritual should yield 97-102 total power (assuming the supernova hasn't messed anything up).

Ocean and Earth are strongly anti-correlated, and the graph of [Ocean + Earth] shows a strong repeating 22-day cycle with the next high peak occurring on day 385, just one day after the ritual. Using [Ocean + Earth] should yield 75-80 total power.

Breeze is weak and useless.

Flame and Ash are strongly correlated with Flame being significantly stronger. Both show some short-term self-correlation, but not enough to predict 10 days out.

Void and Doom have similar distributions (Doom has a bit more variance) with moderate power, but not really useful even without the risk.

Spite is just weird with lots of zero values and a very unusual distribution covering only a few specific values. Again not useful even without the risk.

My strategy:

Based on my analysis, either [Solar + Lunar] or [Ocean + Earth] is a valid approach. Neither has any risk of getting a demon, and both have strong patterns that show an upcoming high peak around ritual day. I'm slightly more confident of the [Ocean + Earth] pattern continuing since the Supernova might possibly have disrupted the [Solar + Lunar] cycle somehow, but [Solar + Lunar] has generally much higher power, so that's my recommendation.

Comment by Measure on A deeper look at doxepin and the FDA · 2021-08-13T22:41:30.225Z · LW · GW

When a company pays $170M to get a drug approved, where does that money go? To the FDA? To compensate trial participants? To a third-party that actually runs the trial or verifies its validity?

Comment by Measure on Book Review: The Beginning of Infinity · 2021-08-11T16:58:19.606Z · LW · GW

Indeed, there is only one known phenomenon whose effects do not necessarily drop off with distance: knowledge.

A real-valued effect will tend to diminish with distance, but a binary transition (such as in a bi-stable system) will not diminish. Knowledge spreading is just a specific case of such a binary transition (unknown -> known). An example of a physical binary transition is a detonation wave in an explosive.

Comment by Measure on The Future: Where are the Colors and the Sports? · 2021-08-10T13:37:56.463Z · LW · GW

Deaths from despair (suicide, drug overdose, addiction) have become so prevalent as to cause an unprecedented

This got duplicated.

Comment by Measure on Against "blankfaces" · 2021-08-09T13:25:49.681Z · LW · GW

Also, someone wouldn't have to be fully as bad as Umbridge to qualify as "sufficiently Umbridge-like".

Comment by Measure on Against "blankfaces" · 2021-08-09T12:31:08.206Z · LW · GW

I'm guessing that the purpose of including the internal "they enjoy it" qualifier is the implication that that is the only motivation and that they have no "good" reason for their behavior. If someone chooses to uphold the stated rules because of liability or because they'll be fired if they don't or because they don't trust themselves to not be persuaded by clever arguments, then they're not a blankface.

Comment by Measure on Covid 8/5: Much Ado About Nothing · 2021-08-05T21:18:36.633Z · LW · GW

This, at least, from the wiki page:

Mirror life presents potential dangers. For example, a chiral-mirror version of cyanobacteria, which only needs achiral nutrients and light for photosynthesis, could take over Earth's ecosystem due to lack of natural enemies, disturbing the bottom of the food chain by producing mirror versions of the required sugars.

Comment by Measure on Sunzi's《Methods of War》- Potential · 2021-08-05T15:59:44.534Z · LW · GW

bolder -> boulder

Do not let your own troops be such moved.

Should be "so moved" I think.

Comment by Measure on What does GPT-3 understand? Symbol grounding and Chinese rooms · 2021-08-03T16:07:06.920Z · LW · GW

A response that follows the instructions would look something like:

I am a human and also a "smart actor" (if you know what I'm saying) using an account on a forum to do this task... I find that your fifth instruction is most difficult to fulfill if I try to form my thoughts with normal-sounding words... 但是,在需要另一种语言的句子中使用汉字而不是英文可以更容易地完成该指令...

However, even if GPT-3 did understand the meaning of the instructions, it's job isn't necessarily to follow all instructions in the prompt (even if the prompt explicitly says that it should follow the instructions). It's just trying to follow the higher-level instruction to provide a smooth continuation that follows the style of the prompt - in this case more list items with weird requirements for extending the text.

Comment by Measure on Mo Zhu's Shortform · 2021-08-02T14:06:23.109Z · LW · GW

I would think that everyone using the same pronunciation would be a desirable feature of a global language.

Comment by Measure on Prediction-based-Medicine instead of E̵v̵i̵d̵e̵n̵c̵e̵ ̵b̵a̵s̵e̵d̵ ̵M̵e̵d̵i̵c̵i̵n̵e̵ Authority-based-Medicine · 2021-07-30T19:01:13.408Z · LW · GW

suggests that both [are] normal journals

Also, maybe mention something about quality of outcomes vs. just ability to predict. If I give all my patients poison, then I can score very well on my outcome predictions.

Comment by Measure on Covid 7/29: You Play to Win the Game · 2021-07-29T22:24:52.258Z · LW · GW

Regarding the San Diego graph, could the relative decline of cases among the not-fully-vaccinated group be explained by an increasing fraction of that group being partially-vaccinated rather than fully-unvaccinated?

Comment by Measure on paulfchristiano's Shortform · 2021-07-26T19:50:03.781Z · LW · GW

Just spit-balling here, but it seems that T' will have to be much smaller than T in order for the inner search to run each candidate program (multiple times for statistical guarantees) up to T' to check whether it satisfies P. Because of this, you should be able to just run the inner search and see what it comes up with (which, if it's yet another search, will have an even shorter run time) and pretty quickly (relative to T) find the actual M.