What are our underlying assumptions about intelligence?
Answers
Answered by
1
The deepest underlying assumption is that intelligence is a physical process or the (emergent) consequence of one, and not a metaphysical process or the (emergent) consequence of one.
That assumption therefore implies that intelligence can be realized in appropriate physical processes, whose nature and characteristics are identifiable and understandable.
AI is therefore, at its base, a dual endeavor:
the endeavor to identify and understand the processes that might or might not lead to inteligence; and, unavoidably,
the endeavor to identify and understand the nature and characteristics of what we call "intelligence".
We're still working hard on both fronts. We're doing significant advances in endeavor 1, but we need a lot more advances in endeavor 2. One problem is that it looks like there's a single example of recognizable intelligence we have available: humans--there are other animals we call intelligent, but they're nowhere near our level of complexity and sophistication.
Having a single, almost totally uniform example of intelligence makes progress in endeavor 2 very slow: it is almost impossible to extrapolate a usable conclusion from a few almost identical exemplars. We've learned more by studying (un)fortunate accidents, damage and developmental irregularities in humans than we have learned by comparative observation of regular-functioning humans. And obtaining more data by purposely damaging humans is very much frowned upon... :-)
Nevertheless, we're slowly getting there.
That assumption therefore implies that intelligence can be realized in appropriate physical processes, whose nature and characteristics are identifiable and understandable.
AI is therefore, at its base, a dual endeavor:
the endeavor to identify and understand the processes that might or might not lead to inteligence; and, unavoidably,
the endeavor to identify and understand the nature and characteristics of what we call "intelligence".
We're still working hard on both fronts. We're doing significant advances in endeavor 1, but we need a lot more advances in endeavor 2. One problem is that it looks like there's a single example of recognizable intelligence we have available: humans--there are other animals we call intelligent, but they're nowhere near our level of complexity and sophistication.
Having a single, almost totally uniform example of intelligence makes progress in endeavor 2 very slow: it is almost impossible to extrapolate a usable conclusion from a few almost identical exemplars. We've learned more by studying (un)fortunate accidents, damage and developmental irregularities in humans than we have learned by comparative observation of regular-functioning humans. And obtaining more data by purposely damaging humans is very much frowned upon... :-)
Nevertheless, we're slowly getting there.
Similar questions