Animal, Vegetable, Mineral - Portable?

Thanks to GameGossip for reprinting a press release announcing Radica's forthcoming portable electronic game called 20Q, licensed from the 20Q.net website. The game seeks to guess an object you're thinking of by asking you 20 questions, starting with "Is it classified as Animal, Vegetable, Mineral, Other or Unknown?" The site is billed as "The neural-net on the Internet", and since it's claimed that "20Q.net is a learning system; the more it is played, the smarter it gets", it'll be interesting to see if Radica's portable version tries to incorporate any learning attributes.

Comments

Half interesting

[Althazzar]

This portable version, even with learning features in it, will still be half as nice as the website version, since the web-version can learn from a lot more tries then the portable version.

Re:Oldest game ever!

[orthogonal]

This game is the oldest game ever.

and not only that, the "learning" isn't anything more adding a discriminator a search tree.

The original used a binary search tree -- each parent question branched to two child nodes, one for "Yes" answers, one for "No" answers.

The version at 20q.net uses a tree where each node has six children (Yes, No, Unknown, Irrelevant, Probably, Doubtful), and makes "guesses" probably based on some accumulation of -- I'm just guessing myself here -- the fuzzier (neither "Yes" nor "No") answers.

So it just classifies human knowledge, and -- big surprise -- it gets "confused" where reasonable people disagree about what attributes the guessed-at object has.

So it's nothing revolutionary technically.

And there's no reason to make it into a single-use portable, given that it could be programmed for any existing portable -- Palm Pilot, Zaurus, Gameboy -- limited only by the size of the database said portable could accommodate. The whole point of Turing machines is that they can be any (programmable) machine -- why this should be a stand-alone, other than because marketing thinks it would sell, I have no idea. (Maybe they can sell stand-alone tic-tac-toe machines too?)

And it's no breakthrough epistemologically: schemes for the classification of all human knowledge have been a hobby-horse of talented zealots at least since the Enlightenment (and come to think of it, wasn't that what St. Thomas Aquinas was up to too?).

Roget's Thesaurus is an example of one of the few classifications of knowledge to actually be useful, but let's not forget various plans by various philosophers to create artificial languages based on "natural" taxonomies of knowledge, or "mathematical" systems encompassing all knowledge, with syntax that would make false statements impossible, and other grandiose plans.

So far, these plans have all foundered on disagreement between reasonable men over what the boundaries and connections between concepts "really" are, and difficulties dealing with different domains of knowledge, to the point that most if not all have had little practical use (Roget's being useful not for its original purpose, an exhaustive classification of everything, but instead as a catalog of synonyms and antonyms best employed by poets and rhetoricians, not scientists or philosophers.)

Of course, just because it's neither new nor particularly noteworthy, does not mean that the US Patent Office might not grant it a patent. But that's another problem altogether.

It's really good

[DarkGreenNight]

After the initial training period it allways guesses what I'm thinking about at the first try!

Are you thinking about Angelina Jolie?

Damn! Here it goes again!

No, I haven't RTFA

I'm not sure how...

[presearch]

but it needs a tie-in with Google.

Re:Oldest game ever!

[iocat]

And of course, if you stump it, you get to add a discriminating question. It's very easy to make up questions that while technically true, don't help the learning aspect. (Think "Is it a turnip?" with the answer "no" for the animal "camel".)