Posts

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

Comments

Comment by Measure on TurnTrout's shortform feed · 2022-01-25T16:05:32.055Z · LW · GW

You also have to assume that the AI knows everything you know which might not be true if it's boxed.

Comment by Measure on Too right to write · 2022-01-21T00:28:32.314Z · LW · GW

Maybe this is just typical-minding, but I would guess that the median person might struggle to say which of faze/phase to put in their own writing but would happily go along with whichever one they encounter on an official-looking website without a second thought.

Comment by Measure on MackGopherSena's Shortform · 2022-01-19T13:24:06.309Z · LW · GW

Is this just postulating that whatever did happen (historically) should have happened (morally)?

Comment by Measure on MackGopherSena's Shortform · 2022-01-18T15:09:48.841Z · LW · GW

I'm confused. What does anthropics have to do with morality?

Comment by Measure on Quinn's Shortform · 2022-01-17T16:47:23.257Z · LW · GW

I'm not sure I understand. If the lever is +100 in world A and -90 in world B, it seems like a good bet if you don't know which world you're in. Or is that what you mean by "acceptably small amount of disvalue"?

Comment by Measure on Personal blogging as self-imposed oppression · 2022-01-16T01:06:38.565Z · LW · GW

You can't make me.

Comment by Measure on Luna Lovegood and the Fidelius Curse - Part 1 · 2022-01-13T14:49:25.944Z · LW · GW

Yep, my bad. I thought that this was still related to the Nine and One-Half thing.

Comment by Measure on Luna Lovegood and the Fidelius Curse - Part 1 · 2022-01-12T14:34:39.433Z · LW · GW

Platform Ten and Three-Quarters

That explains why they didn't get through—they were one and one-quarter too high.

Comment by Measure on Third Time: a better way to work · 2022-01-09T13:41:49.428Z · LW · GW

"Well-earned rest" is the standard idiom.

Comment by Measure on Third Time: a better way to work · 2022-01-09T13:39:33.426Z · LW · GW

Maybe "Earned Breaks"? I did like the 'ratio' pun though.

Comment by Measure on 10 Reasons You’re Lazy About Dating · 2022-01-07T04:38:22.503Z · LW · GW

Dating is for babies

The confusion comes from parsing this as "babies are the ones who should be dating" rather than as "people date in order to make babies".

Comment by Measure on Luna Lovegood and the Fidelius Curse - Part 4 · 2022-01-05T18:16:24.625Z · LW · GW

You mean you don't remember parts 1-3?

Comment by Measure on D𝜋's Spiking Network · 2022-01-04T14:51:21.175Z · LW · GW

For each ⊥i column, if Δ⊥i is in the top 0.66% of those seen before then we perform negative reinforcement.

Do we negatively reinforce False columns that are especially distant from the True column? That seems backward.

Comment by Measure on Open Thread - Jan 2022 [Vote Experiment!] · 2022-01-03T13:29:16.886Z · LW · GW

It's easy to write a comment that's net positive overall. It's hard to write one that's separately net positive on each axis. I expect a system like this would lead to me spending more time crafting my comments and posting fewer (better) comments overall.

Comment by Measure on The Machine that Broke My Heart · 2021-12-30T20:29:51.554Z · LW · GW

I read it as "beep to remind you to log what you're eating".

Comment by Measure on Omicron: My Current Model · 2021-12-29T16:05:24.123Z · LW · GW

I think this only matters if you happen to get an unusually severe case that requires hospitalization, but that still affects the expected value somewhat.

Comment by Measure on A good rational fiction about IT-inspired magic system? · 2021-12-27T23:47:15.133Z · LW · GW

Scott's Anglophysics world could probably become something like this with more development.

Comment by Measure on Worldbuilding exercise: The Highwayverse. · 2021-12-22T20:03:27.194Z · LW · GW

It seems like the equilibrium would push very hard for a very high population density where most people's parents wanted lots of children.

I'm not sure why this is the case?

The average lifetime number of children would have to be exactly replacement at equilibrium, and if people are generally long-lived, this would imply an extremely low number of children per year. This seems implausible since if even a small number of people were willing/able to sustain a high fertility rate, most other people would have to have no children to keep the average low, and the few high-achievers would be the parents of most of the population.

Comment by Measure on DnD.Sci GURPS Evaluation and Ruleset · 2021-12-22T19:50:50.335Z · LW · GW

An empirical-ish strategy of just taking good-looking gear to good-looking areas outperforms most attempts to optimize for anything unless you know a lot about the generating system.

Well, this is basically what I did. I could see some vague patterns, but I wasn't confident in anything and not having a clear picture of the migration patterns made it much harder to analyze the element effects. I did decide to alternate locations on a last-minute whim in case there were any short-term migration patterns, but I probably still had too much repetition.

Comment by Measure on Worldbuilding exercise: The Highwayverse. · 2021-12-22T14:55:43.668Z · LW · GW
  1. Surely there would be some contrarians who try Messing With Time (perhaps they precommit to doing something iff they learn that they didn't do it). What happens to them?

  2. Suppose someone learns how they die (say they confirm it with high accuracy with multiple self-messages and even going to the location to observe the event themselves). Wouldn't this cause them to feel comfortable taking extreme risks at other places/times since they know they'll survive?

  3. It seems like the equilibrium would push very hard for a very high population density where most people's parents wanted lots of children. You mentioned isolated self-parenting pairs. What regulates how many of these exist?

  4. I would expect that a significant fraction of the futile/necessary optimization efforts would be adversarial - analogous to your chess example but on a larger scale with higher stakes, such as political outcomes or group conflicts, with each side trying to out-terminator each other.

Comment by Measure on Six Specializations Makes You World-Class · 2021-12-22T13:23:35.179Z · LW · GW

Related: Being the (Pareto) Best in the World

Comment by Measure on Covid 12/16: On Your Marks · 2021-12-21T00:16:13.009Z · LW · GW

Yep, you want to target your interventions where you'll see the greatest gains relative to no intervention. If you have a miracle cure, you'd focus on those with the lowest untreated odds.

(In the specific case, however, you have to consider whether you're disincentivizing vaccination by giving better treatment to the unvaccinated.)

Comment by Measure on Covid 12/16: On Your Marks · 2021-12-20T20:20:26.915Z · LW · GW

By that logic, you'd rather give it to someone who isn't infected in the first place.

Comment by Measure on Mati_Roy's Shortform · 2021-12-19T13:08:22.769Z · LW · GW

Maybe combine this with a public record of who made which predictions (at least shared with the expert group), so they can converge on their own, Aumann style?

Comment by Measure on Falsifying Ethical Values · 2021-12-19T13:05:02.983Z · LW · GW

two plus two equals one in the Z4 cyclic group.

Zero

Comment by Measure on D&D.Sci GURPS Dec 2021: Hunters of Monsters · 2021-12-17T18:31:05.545Z · LW · GW

In addition to the seasonal pattern of the biologists expeditions as noted by simon and Yonge, they almost always go to three different locations, but occasionally two (once all three) explorers go to the same location. This occurs most often (but not only or always) on weeks that are multiples of 13, which correspond to the exact middle of a "season".

Comment by Measure on D&D.Sci GURPS Dec 2021: Hunters of Monsters · 2021-12-16T17:19:24.993Z · LW · GW

My recommendations:

There are some obvious patterns here, but not enough data to be sure of anything (it's especially hard to figure out what weapons are best against what monsters). Because of this, I'm simply recommending the overall most effective gear and taking a stab at which locations seem to have a good diversity of species at this time of year:

Always take the Toxicala Blowgun and the Electro Chainmail.
Week 6: Thunderwood Peaks
Week 7: The Lordesteppes
Week 8: Devil's Maw
Week 9: The Lordesteppes
Week 10: Scorchsand Shores
Week 11: The Lordesteppes
Week 12: Scorchsand Shores
Week 13: The Lordesteppes
Week 14: Scorchsand Shores
Week 15: The Lordesteppes

Comment by Measure on D&D.Sci GURPS Dec 2021: Hunters of Monsters · 2021-12-13T16:47:39.966Z · LW · GW

The monster knowledge list looks suspiciously like 5 of each genus sorted from smallest to largest.

Several of the monster types have been caught by the hunter in locations where the biologists never found their tracks:

  • Thunderclap Wyvern has only been sighted at The Lordesteppes, but has been hunted often in every location.
  • Raging Windrider has never been sighted, but has been hunted often in Scorchsand Shores and somewhat in Miresmouth Forests.
  • A few others have been hunted occasionally but never sighted.

These seem like good target candidates since they're apparently easy to hunt but would be considered "rare" by the biologists.

Comment by Measure on D&D.Sci GURPS Dec 2021: Hunters of Monsters · 2021-12-13T16:35:27.448Z · LW · GW

I got beaten up and came home with nothing

Which of these does a haul of "None" indicate?

  1. No monster was found.
  2. A monster was found, but it couldn't be beaten and we're not told what monster it was.
  3. One of the above, but we're not told which.
Comment by Measure on [deleted post] 2021-11-29T16:12:52.657Z

I like songs that evoke strong emotions. Here are some of my favorites:

Jonathan Young - The Phantom of the Opera (cover)
Breaking Benjamin - Dear Agony
Disturbed - Inside the Fire
Red - Not Alone
Ad Infinitum - Unstoppable

Comment by Measure on Paxlovid Remains Illegal: 11/24 Update · 2021-11-25T02:44:07.531Z · LW · GW

If the drug hasn't been approved, then those participants weren't going to get it anyway, so they're no worse off in the trial.

Comment by Measure on Morally underdefined situations can be deadly · 2021-11-22T17:32:39.421Z · LW · GW

The point is what it's not obvious whether we'd want an AI to gamble with human extinction in order to avoid morally questionable outcomes, and that this is an important question to get right.

Comment by Measure on [deleted post] 2021-11-22T16:12:15.491Z

In that case, I get a different answer:

The shortest path to the wall avoiding the semicircular FOV is a quarter-circle with length pi/2 times r (arriving at a distance r from the night watch).

Comment by Measure on [deleted post] 2021-11-22T15:15:14.415Z

Please clarify the problem statement (maybe include a diagram of the initial setup):

  1. Does the night watch move? a. If so, what are our relative speeds? b. Does he need to catch me, or just see me and raise the alarm?
  2. Do I understand correctly that the night watch's field of view is a circle with some radius (blocked by the wall)?
  3. Is the night watch right next to the wall, or some distance away?
Comment by Measure on The Meta-Puzzle · 2021-11-22T13:12:55.244Z · LW · GW

A single satanist could say it.

Comment by Measure on The Meta-Puzzle · 2021-11-22T13:04:03.004Z · LW · GW

"I worship God XOR I am married."

Comment by Measure on Split and Commit · 2021-11-22T05:40:30.707Z · LW · GW

Yep, it was a joke saying basically "make sure your alternative isn't a straw hypothesis."

Comment by Measure on Split and Commit · 2021-11-22T01:06:39.718Z · LW · GW

I suppose it's possible that the house elves are to blame. I'll make sure to consider this alongside my other theory.

Comment by Measure on D&D.Sci Dungeoncrawling: The Crown of Command Evaluation & Ruleset · 2021-11-16T18:10:04.590Z · LW · GW

My biggest mistake was modeling the encounters as independent points of failure. I had considered the possibility of hit points or something similar, but I didn't put in the effort to check e.g. how encounter failure rates varied with depth.

Comment by Measure on Improving on the Karma System · 2021-11-15T15:20:38.708Z · LW · GW

You have answered your own question.

Comment by Measure on Improving on the Karma System · 2021-11-15T15:10:43.525Z · LW · GW

I am also perfectly fine with the status quo, but there's still value in experimenting and trying to iterate/improve.

Comment by Measure on D&D.Sci Dungeoncrawling: The Crown of Command · 2021-11-11T17:16:54.133Z · LW · GW

My analysis found that for poison needle traps, clerics and druids were almost as good as rogues, especially at low levels, and the druid will be better for the snake pits.

I admit that I made my choices without considering that later encounters are harder than earlier ones (I suspect this has something to do with lost hp, though I'm still confused by the sharp cutoff in success rates - not only to all parties beat their first encounter, for dungeons with eight or more encounters, the first two encounters are guaranteed wins, and longer dungeons have an even stronger effect.).

Comment by Measure on D&D.Sci Dungeoncrawling: The Crown of Command · 2021-11-09T22:16:42.556Z · LW · GW

Given the potential value of world kingdom domination and the relatively low probability of success, it's probably worth significant risk to borrow/steal additional funds to improve my chances. After all, once I get the Crown, I can repay banish my creditors :)

Comment by Measure on D&D.Sci Dungeoncrawling: The Crown of Command · 2021-11-09T22:00:30.205Z · LW · GW

My party selections:

Lost Temple of Lemarchand: (6kgp - 52%)

Fighter 1
Mage 1
Cleric 3
Druid 1

Infernal Den of Cheliax: (18kgp - 54%)

Fighter 4
Ranger 3
Mage 7
Cleric 4

Goblin Warrens of Khaz-Gorond: (12kgp - 52%)

Fighter 3
Fighter 2
Ranger 3
Cleric 4

Total: (36kgp - 15%)

Comment by Measure on D&D.Sci Dungeoncrawling: The Crown of Command · 2021-11-08T19:06:56.255Z · LW · GW

Deciphering the Unknowns (partial solution):

Lost Temple of Lemarchand - All encounters are known.

Infernal Den of Cheliax - The final encounter of an Infernal Den is 1/2 chance Dragon and 1/2 chance Lich. Of 33 explored, 4/4 with Orcs and 7/7 with Wolves contained a Dragon rather than a Lich (No Infernal Den that contained Orcs or Wolves contained a Lich). The final encounter is a Dragon.

Goblin Warrens of Khaz-Gorond - Each Goblin Warrens uses only a single type of trap from among (Boulder Trap, Poison Needle Trap, Snake Pit). This trap is 1/3 of encounters. The remaining 2/3 of encounters are Goblins. The final encounter is 3/4 chance Goblin Chieftain and 1/4 chance regular Goblins (never a trap). Encounters 3-9 are 2/3 Goblins and 1/3 Boulder Trap, and the final encounter is 3/4 Goblin Chieftain and 1/4 Goblins.

Comment by Measure on What are red flags for Neural Network suffering? · 2021-11-08T14:48:38.097Z · LW · GW

Can you (not) suffer?

Can GPT-3 (not) suffer?

Can GPT-3 feel bad/good?

Are you (not) in pain?

If somebody hurt you, would you be in pain?

If you had X, would you suffer? (different good and bad conditions)

Wouldn't this just tell us whether GTP-3 thinks humans think GPT-3 suffers?

Comment by Measure on Feature Selection · 2021-11-05T21:30:21.582Z · LW · GW

I'm maximizing total reward over the run rather than rate of reward.

Comment by Measure on Y2K: Successful Practice for AI Alignment · 2021-11-05T15:03:14.255Z · LW · GW

They said the death toll could be in the millions, if not hire.

hire -> higher

The feared domino effect of cascade failures in interconnected systems did not happen because the major supply chains contained the largest companies and these co

This quote is cut off.

Comment by Measure on Slider's Shortform · 2021-11-04T17:46:37.493Z · LW · GW

If you just graph fish vs. time, then the one-time gift is a constant function, the steady income is linear, and the "growing promise" stream-of-streams is a quadratic. The fact that a steady income will eventually surpass any one-time gift is because any positive-slope linear function will eventually exceed any constant-value function. Likewise, any 2nd-order polynomial with a positive x^2 term will eventually exceed any linear function. You could keep going with higher order polynomials if you want.

A similar analogy would be a race where even a large head start will eventually be surpassed by a slightly faster car.

Comment by Measure on [deleted post] 2021-11-04T00:59:11.299Z

[The following is in response to a deleted comment.]

You would need rather a large expanse of near-vacuum to avoid destabilizing the planet's orbit as the chaos raced inward to fill the void, and vacuum is quite an ordered state.

I agree with you about the non-consecutive sense of identity, but that should imply that most Boltzmann thought-instants would have random, incoherent memories rather than complete, laws-of-physics-obeying memories. This is, I think, a much stronger argument for a "normal" sequential world than relative size/complexity of brains vs. solar systems, which I still think points in the other direction. I have poor intuitions for these sorts of scales though, so I could easily be mistaken, and maybe there's something I'm not thinking of that would mean even non-consecutive Boltzmann brains would often have coherent memories (maybe the unlikeliness of coherent memories is dwarfed by the relative likeliness of momentary brains).