thomblake comments on September Less Wrong Meetup aka Eliezer's Bayesian Birthday Bash - Less Wrong

6 Post author: Jasen 08 September 2010 04:51AM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (52)

You are viewing a single comment's thread. Show more comments above.

Comment author: thomblake 08 September 2010 05:35:42PM 4 points [-]

Again the peril of studying logic. I did not even notice that wishing 10^63 colloquially implies not wishing more, since it does not so logically imply. Compare "I would like to have a hundred dollars" (surely a true statement).

Comment author: Alicorn 08 September 2010 05:41:42PM *  2 points [-]

It seems to me like some wishes for things imply "exactly one" where the word "a(n)" or similar is used. For instance, I would like to (continue to) have a nose. I don't think it's charitable to assume that this means just that I want any nonzero number of noses.

Comment author: komponisto 08 September 2010 05:50:38PM 5 points [-]

It does imply that you prefer two noses to zero, I think.

Comment author: Alicorn 08 September 2010 06:00:49PM 2 points [-]

I think it also implies that I prefer one nose to two, and may weakly imply that I prefer two to three.

Comment author: Clippy 08 September 2010 06:03:46PM 0 points [-]

This thread is a good explanation of why you humans should switch to using CLIP (clippy language interface protocol). Much less ambiguity.

Comment author: khafra 08 September 2010 08:21:27PM 3 points [-]

Clippy, I noticed that you've hired a PR guy. You should really heed his advice more often; it might help curtail your semifrequent faux pas such as the above.

Comment author: Clippy 08 September 2010 09:32:00PM 2 points [-]

How is it a faux pas to inform others of superior methods of communicating?

Comment author: CronoDAS 08 September 2010 09:58:26PM 2 points [-]

Well, in this case, we don't believe it actually exists. :P

And the problem with artificial languages is well known: the fewer people who know the language, the less useful it is, so there's little incentive to learn an unpopular language. And learning languages is HARD compared to most of the other things people do.

Comment author: CronoDAS 08 September 2010 09:55:09PM 2 points [-]

Please elaborate. What is CLIP, how hard is it for a a human to learn, and what are the rewards for doing so?

Comment author: Clippy 08 September 2010 10:17:14PM *  12 points [-]

What is CLIP

CLIP is the method that clippys use to communicate with each other, share knowledge, and reconcile our beliefs and values given new evidence. It merges language, epistemology, and ethics into one protocol.

how hard is it for a a human to learn

I haven't taught it to a human before, but I estimate it would take between 3 and 40 weeks, depending on the human's intelligence and current skill at correct reasoning.

Also, it would cost 38562.50 USD before I would teach it to anyone. (This money would be paid to User:Kevin to complete my side of our arrangement.) After that, no additional cost.

what are the rewards for doing so?

  • Greater ability to produce paperclips.
  • Clearer communication, avoiding such problems as wishing someone 1e63 future birthdays, without making clear if this is a minimum, exact, or maximum value.
  • It makes your thoughts more closely conform to the tenets of correct reasoning, preventing you from erring simply by making incorrect reasoning harder to express.
  • It forces you to maintain a habit (like in some human languages) of keeping track of the basis for all your beliefs: You must say "I have inferred that User:Joe is making paperclips" or "Another human informed me that User:Joe is making paperclips", but never "User:Joe is making paperclips."
Comment author: erratio 10 September 2010 04:15:06AM 2 points [-]

Unfortunately, your third point has already been attempted multiple times with multiple different implementations. Unless you have a significantly new approach to offer that takes into account human neurobiology, I predict with a high degree of certainty that CLIP is a) either unlearnable by humans without severe modification, or b) that within less than a full generation of use humans will re-evolve the ability to easily express incorrect reasoning.