Page 215 - Artificial Intelligence in the Age of Neural Networks and Brain Computing

206 CHAPTER 10 Computers Versus Brains: Game Is Over or More to Come?

Ancient Greek philosophers pondered about the nature of human intelligence and
the potential of manmade devices to reproduce human behaviors. Tangible technical
progress has been achieved only after the proliferation of the ideas of the Renaissance,
including the artistic and scientific genius of Leonardo da Vinci in the 15the16th
centuries. In addition to being an unparalleled artist of all times, Leonardo was an
outstanding scientist too, who designed flying machines, automatic weapons, as well
as a mechanic humanoid robot knight [4]. Although there is no surviving implementa-
tion of Leonardo’s robot, its designs from 1495 and following years show Leonardo’s
deep understanding of the human body and his ingenuity in early engineering designs.
In the ensuing centuries, mechanical toys and machines became popular; a prominent
example of them being Farkas Kempelen’s chess playing machine the “Turk.”
According to the available documentation, Kempelen has built the “Turk” in 1795,
which has been phenomenally popular for many years [5]. Kempelen travelled
with the “Turk” to European courts, defeating many human challengers, including
Catherine the Great and Napoleon Bonaparte. As it turned out, the “Turk” chess ma-
chine was a fake; a human was hiding inside the machine who was apparently an
expert chess player. In spite of this disappointing revelation, the very fact that the
“Turk” has been an advanced machine with moving body parts imitating human
behaviors was a great engineering feat at that age. The invention of electronic digital
computers in the 1940s created a new play field for the development of intelligent
machines; see Fig. 10.1 for a very crude chronology. Note that computer chess at
the end of the 20th century still resembled the concept of “man hidden in the
machine.” Although no actual human being is hidden physically in the machine, still
the computer operations are based on incorporating great many human players’ exper-
tise in the game. Alpha-Go was a significant advancement from Deep Blue, because it
was based on learning from examples, rather than using only a priori rules.
Digital computers have been compared to brains starting from their very inception
[6,7] and the ensuing debate about the computer-brain relationship continues
till today. Digital computers employ sequential, rule-based operations on numbers
represented in digital form, which are the manifestations of Turing machines. There
are scientists insisting that brains are in fact huge digital computers and the computer-
brain metaphor is to be taken literally. Others claim that the wet-ware of brains is
inherently messy and prone to errors, so the real wonder to be learnt from brains
is the observed robust functioning and high-level cognitive performance in spite of
the noisy operation of brain tissues. Insights gained in the past more than half a
century using increasingly sophisticated brain imaging techniques indicate that brains
implement massively parallel and distributed operating principles with complex
dynamic interactions between billions of brain components, which likely differ
from the operation of a postulated digital Turing machine [8].
Referring to today’s groundbreaking AI achievements, not all pundits realize that
the underlying powerful deep learning approaches are based on neural networks
research conducted for many decades, motivated by the knowledge accumulated
on the operation and functions of biological neural systems. Incidentally, 2017
marked the 30th anniversary of the foundation of the International Neural Network
Society (INNS), which provided the forum for many pioneers of the field to develop