ISSUE385|COMPUTER SHOPPER|MARCH2020 109
ofTennessee,toget some answers.
MacLennan was involved in the
architectural design of the Intel 8086 in
the late1970s,and has had 40 years of
experience in researching and teaching
unconventional computing paradigms.
To the obvious starting point of
whether MacLennan envisaged any
alternative technology replacing
silicon any time soon, we got the
expected response.
“I do not anticipatesilicon-based
digital computation being replaced any
time in the foreseeable future; it will be
difficult to find another technology
that is as flexible and well developed,”
MacLennan asserts.
However,thatdoesn’t mean
remaining with the status quo.
As he explains, “Increases in speed
and density must continue to slow,and
eventually stall, due to fundamental
physical limits. Therefore,Ithink
future computing will involve amix of
computing technologies, both
conventional and unconventional.
“Conventional digital processors will
be complemented by co-processors
designed fordifferent computing
paradigms –for example,neural
network and analogue –and
implemented with non-silicon and/or
non-digital technologies.Unconventional
computing paradigms will also be
used on their own in domains and
environments in which theyhavea
distinct advantage.Therefore,Ithink
the future will bring an expansion of
the range of computing paradigms
and technologies.”
On the question of which
technologies might work alongside
conventional CPUs, MacLennan started
by discussing fundamental principles.
“Todevelop post-Moore’s Law
computing technologies, we will need
to assimilateour ideas of computing
more closely to the physical laws that
implement them. To omany layers of
abstraction between computational
and physical processes limit both
speed and density,” he says.
However,thatdidn’tmean
that he had no particular
technologies in mind.
“Among other
consequences, since the laws
of physics –even quantum
physics –are fundamentally
continuous, analogue
computation has to playa
larger role.This closer
alignment of computation
and physics also means
that we should think offundamental physical processes as
computational primitives, as is done in
quantumcomputing. The physics can
do the computing forus,”henotes.
MacLennan points to massively
parallel analogue computation as
one of the most promising options.
This includes electronic analogue
computation, as well as other
computational media, such as optical.
“I amalso optimistic about quantum
computing, which Iviewas aspecies of
analogue computation, since qubits
represent continuous combinations of
discretestates. One challenge is to
develop ascalable implementation
technology,but several promising
approaches are under investigation,
and Iexpect we will have identified a
scalable technology –analogous to the
integrated circuit –within five to 10
years,”MacLennan adds.
But it’s not all about speed, as his
final prediction makes clear.
“I also see other unconventional
computing paradigms being
used where theyhavedistinct
advantages, such as speed,
power,information density,
adaptability,self-repair,
support fornovel models of
computation and operation in
specialised environments.
Sometimes, speed and density
are not important, and we
maypreferatechnology
that is slower but has much
lower power requirements,”
he concludes.ABOVE:There can
surely be fewer
unlikely ways of
computing than
slime mouldBELOW:Bruce
MacLennan
foresees a
bright future for
unconventional
computing
techniques