7 Outlook
If we consider the future of A.I. certainly we are at a turning point in history where
increasingly we see the practically limitless applications in diverse areas such as
driverless vehicles, medical diagnosis—healthcare, climate prediction, even social
comfort. Our digital computer systems of today are nevertheless being pushed to the
ultimate limits of fabrication with the end of Moore’s Law in clear sight. In other
words, it will become prohibitively expensive to maintain the advances required for
A.I. to create a post-human world. The poor scaling of computer hardware and
software required as combinatorial complexity increases will not be overcome in
future von Neumann machines.
Predictions of increasing computational capacity in comparison to that of the
human brain or the “singularity”, as coined by Ray Kurtzweil are unlikely using
digital approaches. The scenario of computational equivalence to a human was
proposed to give rise to a sudden and massive increase in dominantly
non-biological intelligence, and we propose it can best be approached by using
biological inspiration in making a system that has physical operational characteristics closer to ourselves. The Atom Switch and networks of them have a potential role
to play in Hybrid–CMOS morphic systems where methods such as Reservoir
Computation in physical analogue hardware can be integrated into a morphic system
that utilizes the optimum performances of CMOS digital with the ASN approach.
Already Atomic Switches have successfully integrated into FPGA devices by NEC
reducing energy consumption, footprint, size and transistor count [56]. The Atomic
Switch technologies are more robust in terms of sensitivity to electromagnetic noise
and radiation than Flash making them candidates for robotic and space satellite
applications. The key differences of the ASN approach to conventional computation
are in the elimination of programming and error correcting each step of a calculation
with the RC paradigm. Likewise, ASN devices use distributed fading memory not
RAM similar to living systems. The ability to handle multiple tasks in parallel is
another advantage of such an approach. Although accurate calculations of arithmetic
operations will always be superior in digital systems, analog systems such as the
ASN excel in decision making, or noisy and error prone data that have no precise
solution but rather a range of outcomes with a best guess of the outcome probabilities
a bit like Newtonian vs. Quantum mechanics where deterministic solutions are
replaced by probabilities. The potential impact of A.I. in society has become quite
heated in terms of the dangers it poses to our society and discussions of government
regulation (Elon Musk) and possibly imposing taxes on A.I. robots and systems have
even been proposed (Bill Gates). However, in many advanced countries there is a
future need for such technology as baby-boomers retire and the population of a
young work force declines A.I. will be essential in healthcare, welfare, national
security, and in many other areas of societal enhancements.
Atomic Switch Networks for Neuroarchitectonics: Past, Present, Future
239
If we consider the future of A.I. certainly we are at a turning point in history where
increasingly we see the practically limitless applications in diverse areas such as
driverless vehicles, medical diagnosis—healthcare, climate prediction, even social
comfort. Our digital computer systems of today are nevertheless being pushed to the
ultimate limits of fabrication with the end of Moore’s Law in clear sight. In other
words, it will become prohibitively expensive to maintain the advances required for
A.I. to create a post-human world. The poor scaling of computer hardware and
software required as combinatorial complexity increases will not be overcome in
future von Neumann machines.
Predictions of increasing computational capacity in comparison to that of the
human brain or the “singularity”, as coined by Ray Kurtzweil are unlikely using
digital approaches. The scenario of computational equivalence to a human was
proposed to give rise to a sudden and massive increase in dominantly
non-biological intelligence, and we propose it can best be approached by using
biological inspiration in making a system that has physical operational characteristics closer to ourselves. The Atom Switch and networks of them have a potential role
to play in Hybrid–CMOS morphic systems where methods such as Reservoir
Computation in physical analogue hardware can be integrated into a morphic system
that utilizes the optimum performances of CMOS digital with the ASN approach.
Already Atomic Switches have successfully integrated into FPGA devices by NEC
reducing energy consumption, footprint, size and transistor count [56]. The Atomic
Switch technologies are more robust in terms of sensitivity to electromagnetic noise
and radiation than Flash making them candidates for robotic and space satellite
applications. The key differences of the ASN approach to conventional computation
are in the elimination of programming and error correcting each step of a calculation
with the RC paradigm. Likewise, ASN devices use distributed fading memory not
RAM similar to living systems. The ability to handle multiple tasks in parallel is
another advantage of such an approach. Although accurate calculations of arithmetic
operations will always be superior in digital systems, analog systems such as the
ASN excel in decision making, or noisy and error prone data that have no precise
solution but rather a range of outcomes with a best guess of the outcome probabilities
a bit like Newtonian vs. Quantum mechanics where deterministic solutions are
replaced by probabilities. The potential impact of A.I. in society has become quite
heated in terms of the dangers it poses to our society and discussions of government
regulation (Elon Musk) and possibly imposing taxes on A.I. robots and systems have
even been proposed (Bill Gates). However, in many advanced countries there is a
future need for such technology as baby-boomers retire and the population of a
young work force declines A.I. will be essential in healthcare, welfare, national
security, and in many other areas of societal enhancements.
Atomic Switch Networks for Neuroarchitectonics: Past, Present, Future
239