from Hacker News

Stanford AI Class Lecture Notes

by ziyadb on 10/10/11, 12:38 AM with 22 comments

• by xarien on 10/10/11, 1:14 AM

  Google still has a very long way doing an adequate translation between a language such as Chinese and English. I'd be surprised if Google was taking such a pedestrian approach in solving that particular problem.

  As it stands, languages such as Chinese are intrinsically implicit in nature. In fact, the more adept at the language, the more you can express with less. If you follow the literature back a couple thousand years, the amount expressed in a few characters is absolutely astounding.

  If you take the example they use at the bottom regarding wonton, it's down right criminal to map the grammar in such a hurried manner. For one, just from the romanization of wonton, the AI should be able to gauge that it's looking for 2 characters and not 1 (1 character per syllable). However, in the case of the menu, the wonton egg drop soup drops a character to save some space.

  Taking a straight forward CFG approach will never result in an accurate translation. What may work well is to do multi-pass contextual analytic processing in parallel.

• by guscost on 10/10/11, 7:38 AM

  Couldn't this understanding of intelligence limit the ways in which we can describe and emulate the activity of intelligent creatures?

  For example, the interfaces and processors are all very clearly defined and separated in those diagrams. Unfortunately, natural intelligence does not seem to work in the same way. The inputs to a real human do not get processed in the same places, even when they might be coming from the same sensor. Obviously the patellar reflex doesn't make it past the spinal cord, and I've never actually believed that the spectrum of intelligent behaviors can be sorted into "conscious" or "unconscious" categories, by including some sort of wet Boolean or whatever.

  We could think of the brain's implementation as the sum of its internal and external interfaces, but how the hell would we model that without involving unreasonable error margins?

• by bstar77 on 10/10/11, 11:30 AM

  This is extremely frustrating, has the login been removed from ai-class.com? I can't get the videos to work (except the intro) and I can't find anywhere to login. I've tried Chrome and FF.

  Update: Now the videos are saying I need Flash 9 (intro previously worked). Bizarre. I just went to youtube to watch the videos, unfortunately they are not organized well or queued so searching for them is a pain.

  This page at least has them all easily accessible: http://www.youtube.com/user/knowitvideos#p/u

• by vimalg2 on 10/10/11, 3:49 AM

  Exported to PDF and re-uploaded for convenience.

  http://dl.dropbox.com/u/5137/pdf-link/StanfordAI-UnitOne.pdf

• by jcarden on 10/10/11, 2:08 AM

  LOL, here's what I just got when I tried to log in

  'Ooops

  Our servers are off having a quick coffee break. Wait a second and refresh the page. If you still get this message, we apologize and ask that you try again a little later.'

• by AlexC04 on 10/10/11, 1:33 AM

  I thought they were going to be posting videos of the lectures... but when I logged in today and couldn't see the it anywhere. Did I misunderstand something? Is it posted? If yes, where?

• by xarien on 10/10/11, 1:16 AM

  Actually come to think of it, I bet IBM's G2 system would work amazingly well with translation.

• by singh on 10/10/11, 5:43 AM

  I too would like to have discussions like xarian's thoughts for all lectures.

  Regarding the lecture notes - is there a wiki where we can all contribute to? Earlier today, the google doc was complaining about too many people editing the document.

• by rottendoubt on 10/10/11, 8:17 AM

  This is great. Thanks for sharing and please keep it coming! =D