iq for ai?

they're talking about an intelligence quotient (iq) test for artificial intelligence (ai) systems.

granted, alan turing pointed out this need (with the turing test) a long time ago.

nonetheless, a good idea.

now, the thing we want to make sure of is that the test isn't worthless. in other words, we wanna take our time and make sure that the test gives us relevant, useful, accurate information, and that we truly _understand_ what the results we are given _actually mean_.

if we (and when i say 'we' what i really mean is whoever ends up making this thing (work)) can accomplish these two things, then the resulting test might actually be a wonderfully useful tool.

rather than just raw tech specs (which, sadly, not enough people actually understand anyway) a user could get a human-readable list of functions or categories of functions (which will be easily understood to project _how_ this new device or program will integrate into their daily lives--useful information).

what this would accomplish is the same it would accomplish in any other field or application--it creates a standard.

and a _good_ (get me here, i'm talking a GOOD) standard can spark innovation like _no_ amount of proprietary technology can.

case in point--the modular desktop pc (the oldschool ibm and it's clones). that open standard started the revolution that _BUILT_ the computing industry as we know and think about it today.

so, i guess i'll say it once more--this will be cool (and good, and useful, and wonderful) if and only if....they do it right.

--cid

0wn yourself