INFORMATION TECHNOLOGY SYMPOSIUM
193
spreadsheet in a billionth of a second, but that doesn't give you the sub–
tlety, the richness, the depth, and the complexity of human intelligence.
How are we going
to
achieve the software of intelligence? I think the
most compelling way is
to
use an entity that exhibits those qualities–
the human brain. We have a few dozen in this room, and it's not hidden
from us: the organization of the human brain, how it processes
information, its knowledge, and its learning are available if we care
to
take a look at it.
We're well down the path of reverse-engineering the human brain by
examining it in detail and understanding its methods. We've already
been able to examine some of the regions of the brain, which is not a
tabula rasa.
It's several hundred different specialized regions that are
organized differently; each has a different way of processing informa–
tion. We actually have an understanding of several of them. The speech
recognition software we've developed, for example, performs the same
transformation of auditory information that the early auditory cortex
does, and that allows the software to understand human speech with
much greater accuracy. We're beginning to understand the methods in
those regions where we have some access. So, if we could understand the
human brain and have a complete understanding of how it works, we
would be able to use those insights to build intelligent machines.
How are we going to get access
to
that information? Well, we already
have brain scanning that can look non-invasively at the human brain;
and the speed, resolution, and bandwidth of brain scanning is acceler–
ating along with the other technical processes. One scenario I find par–
ticularly compelling is to scan the brain from inside. We already have
scanning technology where if you put the scanning tip-the tiny pin–
sized scanning tip-in physical proximity to the neural features, you
can scan the brain with extremely high resolution and see the
interneural connections, even the neurotransmitter concentrations,
where our learning resides.
If
you ask "Where's my knowledge of any
particular field?" or, "Where are my memories in learning?" it's a vast,
extremely complex, distributed pattern in particular regions of the brain
made up of interneural connections and neurotransmitter concentra–
tions, synaptic clefts, and other neural components. We can see those,
and we actually have technology we can touch and feel today to scan
those neural features with extremely high resolution-if the scanning
tip is in physical proximity to the neural features being scanned.
So, in order to scan the entire human brain, we simply take the scan–
ning tip and move it to every single position in the brain and take a high–
resolution picture of it. How are we going to do that without making a
