[ExI] Augmentations to Science

[Ben]

Will Steinberg <steinberg.will at gmail.com> asked:

"Do any of you have any idea what a *very* rough computerized insight 
model might look like?"


You need to define 'insight' first. I suspect different people will have 
different definitions.

To me, it's pretty much the same thing as 'gut feeling'. In other words, 
a cognitive shortcut, where some idea 'feels' right even though you 
can't see immediately why it should.

I think this is probably an 'unconscious analogy' type of thing, where 
you've detected an analogy to something else you've experienced in the 
past, but aren't consciously aware of what the something else is, and 
this process is tied to an emotional response, making it jump out more 
quickly than the normal micro-evolution process that goes on in the 
cortex when we're thinking.

The problem with insights is they're sometimes wrong.
Any algorithm that captured the process would probably have to build on 
other algorithms that model thinking anyway, and these would probably be 
fast enough that there'd be no point in having an additional 'insight' 
system which is less reliable. For a biological organism, the 
evolutionary benefits of the extra speed probably compensates for it 
being wrong a certain amount of the time. For a non-biological thinking 
agent, that won't apply.


Ben Zaiboc