Lumifer comments on Open Thread, January 4-10, 2016 - Less Wrong

5 Post author: polymathwannabe 04 January 2016 01:06PM

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

Comments (430)

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

Comment author: Lumifer 05 January 2016 06:48:11PM 1 point [-]

narrow AIs that can supposedly identify an author from their writing

It's not a "narrow AI", it's a straightforward statistical model. Or, if you prefer, an outcome of machine learning applied to texts of different authors.

identify sockpuppets

A voting sockpuppet doesn't post except to get the initial karma. It just up- and down-votes -- there is no text to analyse.

Comment author: gjm 05 January 2016 08:52:20PM 0 points [-]

There's a pattern of voting to analyse, though...

Comment author: Lumifer 05 January 2016 08:57:54PM 0 points [-]

Sure, but that's a much easier problem.

Comment author: gjm 05 January 2016 10:13:06PM 0 points [-]

I'm not convinced it actually is. (To get useful information out of, that is.)

Comment author: gilch 05 January 2016 07:24:20PM *  0 points [-]

It's not a "narrow AI"

https://en.wikipedia.org/wiki/Weak_AI It is by that definition. Of course, words are only useful if people understand them. I know LW has some non-standard terminology. Point me to the definitions agreed upon by this community and I'll update accordingly.

A voting sockpuppet doesn't post except to get the initial karma.

Sounds like the initial karma threshold is too low. I have various other ideas about how to fix the karma system, but perhaps I should hold off on proposing (more) solutions before we've discussed the problem thoroughly. If that's already been started I should probably continue from there, otherwise, do you think this issue (karma problems) merits a top-level discussion post?

Comment author: Lumifer 05 January 2016 08:12:18PM -2 points [-]

https://en.wikipedia.org/wiki/Weak_AI It is by that definition.

I still don't think so. "AI" is a very fuzzy term (the meaning of which changes with time, too) but in this case what you have is a fairly plain-vanilla classifier which I see no reason to anoint with the "intelligence" title.

do you think this issue (karma problems) merits a top-level discussion post?

Karma has been extensively (and fruitlessly) talked about here. If you want to write a top-level post about your proposals, it might be a good idea to acquaint yourself with the previous discussions here (as well as the experience of other forums, from Slashdot to Reddit).