* [...] In that case, a heuristic might be used to more [efficiently arrives] at an option that near-optimally satisfies the underlying goal [...]
This is a substantially simplified and lazy answer by the way. :-D As for sampling patterns, that depends on the type of algorithm used to search for a good enough solution. There are usually ways to accelerate that where more than one option can be discarded at once (ex: binary search).A very simple example of this type of rapid filtering with ice cream flavors is this: say the robot samples chocolate (or has done in the past and stored its assessment in memory). Then it might be able to determine that any variation of chocolate does not satisfy its underlying objective whatsoever, so it might be able to discount multiple flavors at once involving chocolate. This allows it to quickly filter out all options involving chocolate.
There are almost always time/resource constraints on making decisions. An algorithm that requires a few hundred years to arrive at a decision is probably not going to be very applicable. So there's almost always a pressing need to be able to arrive at a decision in a reasonably efficient way which constraints the time and resources available. This may also limit the machine's ability to arrive at the most optimal solution and instead require settling for "near-optimal", or even just "adequate/satisfactory".
Additionally time constraints are not specified. How much time it can spend on the problem also widely affects the algorithm used to a certain extent.
Scroll Down to Read Other Opinions
Lmao. That's weird, JJ.
@Jamie05rhs Yup, that's a girl robot only I wish her nipples were not showing cause it's a little over the top...
@VIVANT that was meant to be a joke. Hopefully you took it that way...