Wednesday, January 21, 2009

Which of the following can be done at present?

Which of the following can be done at present?

- Play a decent game of table tennis
- Drive along a curving mountain road
- Drive in the center of Cairo
- Play a decent game of bridge
- Discover and prove a new mathematical theorem
- Write an intentionally funny story
- Give competent legal advice in a specialized area of law
- Translate spoken English into spoken Swedish in real time
- Playing an (entertaining) game of Soccer
- Solving NYT crossword puzzles at close to expert level
- Navigating in deep space
- Learning patterns in databases (datamining…)
- Supporting supply-chain management decisions at fortune-500 companies
- Learning common sense from the web
- Navigating desert roads
- Navigating urban roads
- Bluffing humans in Poker..

Thoughts anyone? :)

9 comments:

  1. - Drive along a curving mountain road
    - Drive in the center of Cairo
    - Play a decent game of bridge
    - Write an intentionally funny story (put the right words in and it could be funny, at least to one person :S)
    - Solving NYT crossword puzzles at close to expert level
    - Learning patterns in databases (datamining…)
    - Navigating desert roads
    - Navigating urban roads
    - Bluffing humans in Poker..
    - Supporting supply-chain management decisions at fortune-500 companies
    - Playing an (entertaining) game of Soccer (with itself seen here >> http://www.youtube.com/watch?v=fjSvDUNamtc)

    ReplyDelete
  2. Bluffing some humans in poker may be popular, but not most, and certainly not the good ones. Why?

    1) Empirically, online gambling would be worthless if only bots won.

    2) More rationally, for a bot to beat a human in poker, it would seem to have to simulate the complex social reasoning humans go through when attempting to guess whether or not someone else is bluffing. (Planning and plan recognition in an awkward and extremely social situation)


    Solving crossword puzzles? I’m not an expert or pro at them, but I imagine a combination of context and related words (á la Google Sets), definitions, and brute-force techniques would make solving such puzzles almost trivial.

    ReplyDelete
  3. I saw a furi project where they were using facial recognition and body positioning to determine the mood of the person and then put a reaction to the mood made by the person, maybe that can be used? not sure though if it will be effective.

    ReplyDelete
  4. Just some thoughts:

    I think it's worth noting that several of these bullets have subjective terms that are open to interpretation.

    "decent game of table tennis"
    "decent game of bridge"
    "funny story"
    ...

    For example, if I find cats to be funny, then a story involving a cat headbutting a street light to try to get the pigeons perched atop to fall down would most likely be considered funny. The same story might not be funny to anyone else. Similarly, if I'm terrible at bridge (haven't played in years =D ) then a decent game of bridge to me might be an awful game to another, more experienced individual.

    Moving on, one of these bullets really stood out as interesting to me: "Discover and prove a new mathematical theorem." I'm just starting to learn about AI, so when asked if it is this possible at present, I can't yet give a definite (or correct) answer. However, the concept of feeding a system enough information such that it can begin to create new theorem seems to be what acting rationally is all about. Just think of the things we could discover (or disprove!) with an intelligent agent program capable of this.

    To add to the poker bluffing comments, I would imagine that bluffing a human is quite easy. Bet strong hands and fold weak hands 95% of the time, and then switch it up 5% of the time. However, playing a solid game of poker by reading and reacting appropriately to signals would be much more difficult, as Alan mentioned.

    ReplyDelete
  5. As a follow up to my post, I just noticed slide 17 (hidden slide? whoops) that says:

    1996: EQP proves that
    Robbin’s Algebras are all boolean

    So I guess it IS possible.

    ReplyDelete
  6. - Play a decent game of table tennis

    Playing a decent game of table tennis could be interpreted in many different ways. To me, watching or playing a sport is only fun when there is a possibility that someone will mess up. Playing a bot at table tennis would be quite tedious if the bot has perfect calculation of speed, strength, and accuracy with each hit. It would be quite difficult to adjust the bot to play like a human, which plays according to real time thought process and emotions would not be easy.

    ReplyDelete
  7. The selection I find most interesting is "Learning common sense from the web". Given that the web is littered with junk, viruses, false information, spam, pornography, and the like, it seems nearly impossible for a HUMAN to gain accurate knowledge from that source, much less a computer. Still, with correct human input and aid, the web could be a fast and resourceful way to learn common sense if it were harnessed correctly. At least, that's what I think... I am not very knowledgable on the limitations of A.I.

    ReplyDelete
  8. I can see the New York Times crossword puzzle solver being done by a bot. There seems to be a common set of answers with similar clues which could be brute force programmed. If the number of available blanks for each answer was also known, the program could then draw from this repository in attempts to solve future crossword puzzles at a seemingly expert level.

    ReplyDelete
  9. "Translate spoken English into spoken Swedish in real time"

    This can certainly not be done now, nor do I think that it is possible (in real time). Though, if you have seen the Iron Man movie, they show a real time translation on a computer towards the end of the movie. I laughed at that part.

    ReplyDelete

Note: Only a member of this blog may post a comment.