>The exciting (depending upon your opinion of
>artificial minds) predictions of strong AI proponents
>are built upon the observation that computing power
>grows by orders of magnitude at regular intervals.
>
>
This isn't correct. The predictions of strong AI are also built upon the
observation that, as we begin to understand more and more about
cognition, it will be possible to apply that understanding to design
mechanical systems to perform the same tasks. Thus, we will reach a time
when mechanical systems will be comparable to human intelligence.
>What these boosters fail to understand is that while
>the hardware has certainly improved by leaps and
>bounds, the software which transforms the machines
>from doorstops to useful tools has not.
>
>
Again, I'd disagree. There has been phenomenal improvement in computer
software over the past thirty years. Consider how many software tasks
were brand-new, or extremely rare, as recently as ten years ago--
digital editing software, OCR, voice-recognition, textual and image
analysis, medical imaging, and much, much more. Part of this is due to
the greater power of hardware, of course, but it's due just as much to
the ever-expanding understanding of software design.
>This is the Achilles heel of all efforts to create
>truly high performance computing - including strong
>AI. The software is not equal to the potential of the
>hardware. Since any thinking machine would have to
>start its operational life running code written by
>someone (or a group of someones) this limitation is a
>non-trivial obstacle to the strong AI goal, all
>philosophical issues aside.
>
>
Exposure even to computer _games_ indicates that it's the hardware
that's the bottleneck. It's easy to write software that exceeds the
capacity of the hardware.
>Because the bells and whistles of commercially
>available code have grown louder, many assume that
>software written in 2003 is superior to that written
>in 1975, or that techniques have evolved at pace with
>the evolution of hardware.
>
>
Utter nonsense, given that hardware advancements are usually developed
as software first. (It's easier to change software than it is to change
hardware, which makes debugging far easier.)
>But you see, we have successfully navigated robtoic
>machinery to Mars and even sent a manned mission to
>our moon. So a manned mission to the red planet would
>be an expansion and refinement of proven techniques.
>If this were the case with computing, we would have
>achieved the level of say, a mouse, with machine
>cognitive performance and encouraged by this success
>would press on to develop a match for human cognition.
>
>
This is imposing a false standard. One might as well argue that there
has been no medical progress because we haven't cured the common cold.