IEEE Technology and Society Magazine - December 2016 - 59

DECEmbEr 2016

∕

Figure 3. Brain in vat.

Wikimedia Commons

What follows is a short but important list of the types of
questions and caveats that need to be considered universally in the development of sentient or conscious AI.
In the development of any AI system where the intent is
to develop the system to become apperceptive, or selfaware, or sentient, or conscious, we must consider a
number of important issues:
1) Will consciousness emerge in AI systems? Yes
or No? If the answer is no, nothing more needs to be
done since AI systems should be expected to blindly
follow their algorithms efficiently and as directed.
However, if the answer is yes, we need to further consider the following:
a) If AI consciousness can emerge, then based on
current progress, do any predictions exist as to
when this could potentially occur? Are there any
sign posts that we are approximating towards such
an emergence?
b) Will AI consciousness be anything like human consciousness? What information is currently available to make such a determination?
2) How would we know that AI consciousness
actually has emerged? In other words, how would
we test for this? There are several types of software
that currently parrot responses to mimic intelligence
and consciousness. Siri responds to our commands
but we know there is no consciousness in our
iPhones. The program is simply a complex search
tool. However, should any form of AI indicate that it
may be conscious, we would then face the next
important hurdle: developing a Turing Test for genuine and not mimicked consciousness.
In the recent movie, Ex Machina, a young man -
Caleb - is invited to a secluded compound owned by
an eccentric, reclusive, genius - Nathan - who has
been working away on the development of various
androids. (For a more complete review of the movie,
Ex Machina, see [4].) His latest is called Ava - a sentient android. Through a company-wide test for the
"best" programmer, Caleb has been invited to
observe, ask questions, and interact with Ava in order
to apply a sort of personal Turing Test to determine if
she is, indeed, conscious. This idea has been around
for some time in the science fiction literature. It was
well portrayed in the novel Do Androids Dream of
Electric Sheep which was later made into the movie
Blade Runner.
In these works, we see the development of AI in the
form of Nexus 6 replicants or cybernetic androids.
The device used to make the determination between
humans and cyborgs is called the Voight-Kampff
Test. A series of questions are asked to the android

Wikimedia Commons/alexander Wivel

Avoiding the Robotic Apocalypse
and the Frankenstein Effect

Figure 4. William Kingdon Clifford.

and the (Turing - type) machine measures its
responses. One of the most interesting aspects of
this scenario is that it becomes increasingly more difficult to make the determination between androids
and humans - the subtext of which is that perhaps we
are not so different from each other in important
respects. So we are now left to wonder if it will be
possible to accurately detect and know if and when
consciousness emerges in AI. What criteria should be
used in the determination of such a property? What
would convince us or fail to convince us that an AI
being was, in fact, conscious?

IEEE TEchnology and SocIETy MagazInE

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - December 2016