on the rationalist foundations of AI

[gusl]

Ok, I can be too much of a rationalist to live my life properly, but does this make me a bad scientist/philosopher?

Winograd, for one, has been critical of rationalistic (reductionistic?) tendencies in AI.

from William J Clancy - Practice Cannot be Reduced to Theory: Knowledge, Representations, and Change in the Workplace:

Fig. 1. Rationalist view of knowledge and representations

These seem like the philosophical foundation for my views about AI & intelligence in general.
Since they've nailed down my view, I guess I really should read Winograd to see if they can change my mind.


Btw, let me hypothesize that "neat" vs "scruffy" orientation correlates with political opinion.


Right-Wing AI:
* rationalist
* truth-conservative

Left-Wing AI:
* evolving
* non-monotonic
* connectionist
* "situated cognition"/"embodied cognition"

Comments

[bram.livejournal.com]

Is Winograd the one who wrote the Shrdlu (sp?) program that moved blocks around? Ahh, it's harder to show of dimly remembered facts when Google knows all.

That tree rubs me the wrong way. It seems to shoehorn reality into the easiest ways it can be represented by computers. Real human memories are much more complex than von Neumann style computer memories. I.e. each time you remember something, the memory changes, the same memory can be stored in different parts of the brain, etc. And in terms of physics, I would disagree that information consists of "features" or that there are "objects".

[bram.livejournal.com]

That's "show ofF" and not "show of".