computers still lack the
fluidity, adaptability, open-endedness, creativity, purposefulness,
and insightfulness we associate with the supreme
achievements of human cognitive ability.


I am guessing that few would doubt
they also still lack the fluidity, adaptability, creativity, purposefulness,
and insightfulness we associate with the
supreme achievements of human cognitive ability.






Is it still true that people are smarter than machines?
And if so: Why?


While I do not think anyone would claim
human-like performance has yet been achieved, I am
actually optimistic that incremental progress is occurring
in all of these areas, at least up to a point.


computers remain, for now, fundamentally nothing
more than tools in the hands of their human designers and
users, and not autonomous, independent, self-directed,
thinking beings, like people.



More computer power might be helpful, but it seems
pretty clear that this alone will not be sufficient.


It may well be, then, that over the next decade, the
butterfly will finally emerge from the chrysalis, and truly
parallel computing will take flight.



It should be noted that the issues here are far from
trivial. It is very difficult to know exactly how to frame the
computational problem. To underscore this point, consider
a cognitive system faced with a series of situation.consequence
observations in some domain, and let us assume we
all agree that it would be a good thing if the system could
use the data to learn something about the relationship
between situations and consequences. How best should we
construe what should be learned in this situation? Currently
in the field of cognitive science, there are two views on this
question. One holds that we should construe the learnerfs
goal as one of extracting a structured statistical model of
the environment.one that explicitly attempts to find the
best type of structure to represent the data, and within this
the best instance of a structure of a given type [9]. An
alternative to this, however, is the position that any taxonomy
of alternative types will always provide at best only
an approximate characterization of natural structure, so that
it is better to define the goal more directly in terms of the
problem of optimal prediction, allowing the internal model
to remain inchoate instead of explicit (as in a neural network
representation; see [17], for further discussion).


17. Rogers TT, McClelland JL. Precis of semantic cognition, a parallel
distributed processing approach. Behav Brain Sci.
2008;31:689.749.




Human mental abilities are profoundly
shaped by experience, and that experience is
structured by social, cultural, and governmental institutions.
Even in the first few months of life, when the child is nurtured
primarily in the informal social and cultural context of the
immediate family, many important changes occur in the
childfs cognitive, social, emotional, and linguistic capacities
that are crucially dependent on the childfs experience. The
effort to understand how human cognitive abilities arise will
depend heavily on taking full account of these influences,
and success in achieving true human-like intelligence in
artificial systems may rely on the creation of systems that can
exploit these influences (see [25]).



25. Weng J, McClelland JL, Pentland A, Sporns O, Stockman I, Sur
M, et al. 
Autonomous mental development by robots and animals.
Science. 2001;291:599.600.