Contest for outlining rules for this contest.
post by abramdemski · 2022-02-21T18:44:43.990Z · LW · GW · 11 commentsContents
12 comments
Each round of the contest will produce the rules governing the next round. Any more information would be too much of a hint. This is dumb and not worth spending time on; I nonetheless look forward to any submissions.
11 comments
Comments sorted by top scores.
comment by Daniel Kokotajlo (daniel-kokotajlo) · 2022-02-22T03:04:56.838Z · LW(p) · GW(p)
I think this is a comment section that could benefit from the distinction between upvotes and agreement.
comment by Jackson Wagner · 2022-02-21T20:48:37.746Z · LW(p) · GW(p)
The rule for the next round of the contest is:
- The most-upvoted suggestion will become the rules for the next round of the contest, subject to the constraint that:
- The winning suggestion for the next round must describe a contest that an ordinary person could reasonably enter with less than a day's effort; eg asking for people to write and submit a google doc rather than asking contestants to create a 1000-acre wildlife preserve or implement a set of rules that is clearly paradoxical/impossible.
comment by Vanessa Kosoy (vanessa-kosoy) · 2022-02-22T14:06:25.357Z · LW(p) · GW(p)
In the next round, we will consider 3 plans:
- Plan A: The winner of the -th round is either the most upvoted comment with even karma, or the most upvoted comment with odd karma, according to whether the Turing machine with Gödel number halts or not.
- Plan B: The rules for the next rounds will be determined by the system for which we can prove the best lower bound on the counterfactual expected amount of happiness in the universe it would produce.
- Plan C: We will generate rule systems by sampling 10,000-character long texts randomly until we encounter a proposal that violates the Geneva convention.
We will then construct a blockchain-based cryptocurrency prediction market, and select the plan that, according to the market, produces the maximal expected number of paperclips.
Also, if my proposal wins, on the next round we will build a superintelligent AI that will mock anyone who didn't vote for my proposal on this round. The mockery will recur on April the 1st of every year for a Graham's number of years.
comment by Richard_Ngo (ricraz) · 2022-02-21T19:30:11.413Z · LW(p) · GW(p)
The rule for the next round of the contest is: the entry which wins must be a copy of this sentence with the ordinal incremented, followed by the 1st letter of Abram's best explanation of the thing he just said he doesn't want to hint about.
Replies from: daniel-kokotajlo↑ comment by Daniel Kokotajlo (daniel-kokotajlo) · 2022-02-22T03:18:15.090Z · LW(p) · GW(p)
Below follows my speculation about what Abram might be thinking.
My wild guess is that Abram has been thinking again about some ideas he mentioned to me years ago -- stuff having to do with how we need to be able to build (or define) a system for thinking/learning that can legislate for itself rules for how to think, to avoid various molochian problems that plague e.g. logical inductors and decision markets. But the legislation needs to be sensible somehow. You need some critical mass of wisdom already to know which rules to impose. Can we construct a "seed," some minimal starting point that will eventually bootstrap its way to a wise agent? Or is the simplest such seed insanely complicated? Going further out on a limb here, I'm guessing that while thinking about this stuff, he thought of the idea for the contest--the commenters are like traders in the market, they are collectively pretty dumb/unwise at least at first, they have this great power to legislate rules for the future... will they be able to bootstrap their way to something competent and wise? Probably not but it's a fun game anyway.
Abram, have you heard of the game Nomic? You are basically proposing Nomic here.
comment by Yair Halberstadt (yair-halberstadt) · 2022-02-22T10:48:21.462Z · LW(p) · GW(p)
The rule for the next round of the contest is:
The most downvoted suggestion will become the rules for the next round.
Not a sensible idea, but I would love to see how people respond - do they downvote good ideas? Do they deliberately suggest bad rules so they get downvoted?
comment by Quintin Pope (quintin-pope) · 2022-02-21T21:44:04.390Z · LW(p) · GW(p)
Each round of the contest will produce the rules governing the next round.
comment by tailcalled · 2022-02-21T20:34:39.629Z · LW(p) · GW(p)
Each submission in each round should contain a link to a form made in e.g. Google Forms (or whichever form system you prefer). The form should have 1 to 6 questions of interest to the submitter, as well as a question asking for a unique identifier of the respondents so the data for different respondents can be linked across the forms. The data for each round should be published for anyone to analyze.
comment by Charlie Steiner · 2022-02-23T00:32:42.876Z · LW(p) · GW(p)
If someone manages to lock themselves in as winner in perpetuity, they win the contest.
comment by SarahSrinivasan (GuySrinivasan) · 2022-02-22T00:04:58.495Z · LW(p) · GW(p)
Exactly 7 days after contest entries are posted, note the net upvotes of each top-level comment. Only top-level comments with net upvotes higher than the average and less than twice the average contribute towards rules for future rounds.
comment by devansh (dpandey) · 2022-02-21T20:17:19.416Z · LW(p) · GW(p)
Every submission must be a 26-letter combination of random lowercase letters with no spaces. The entry that is closest to a randomly generated submission wins.