Artificial intelligence
is poised to be one of the last greatest and riskiest innovation that will ever
be made by humanity. Computers are evolving at a rapid pace, and this has made
us aware that our position in the hierarchy of lifeforms is not a permanent
one. The result of this awareness has been uproar,
mostly among intellectuals, who are continually finding the task of guiding the
rest of us through the binary coded maze of the 21st century complicated. We don’t yet know how the
development of AI will turn out, but two scenarios have been proposed as the
most likely. Scenario number one involves the man versus machine paradigm, where
we oppose but eventually submit to our computer overlords. The second scenario is based on the man plus machine model where we
help AI help us, humanity happily coexists with superintelligent machines and
possibly upgrades itself into one. We can all agree that scenario two is the
most desirable. However, the difficulty
lies in making it the most probable. At the moment, both situations can play out.
The probability that
humanity will be engaged in a losing battle with AI is pretty high because of
our stupidity and insecurity. Ignorance is the primary reason this is likely to
happen. Nick Bostrom, who has warned that AI might lead to our extinction,
compares our current situation to that of a child playing with a grenade (Bostrom). We might be the
ones developing the AI, but that does not imply we have a clue what we’re
doing. We might giggle as we pull the
pin, but we’d be blown to bits before we realize what happened. Our advantage
is that we have been aware of our knowledge limit, which is primarily why the
most significant branch of AI development is machine learning. Computers can learn
because they are now capable of listening and seeing, which makes independent
knowledge acquisition possible. However, we have no way of telling the kind of
power that we have handed machines, and this is what makes AI an existential
risk (Bostrom).
Another reason we might
form an antagonistic relationship with AI is
linked to our insecurity. Via machine learning, humans will become economically
expendable. Many things that we do will be done better by robots, and future
businesses will optimize through full-scale automation. This doesn’t augur well for humans since many of us will be left
jobless, and without jobs, many capitalist
souls will lack a raison d'être. By the time we
reach the halfway mark of this century, we are likely to witness the
emergence of a useless class (Harari). A social class that makes no economic or artistic
contribution to the economy. This
is likely to happen because it is becoming increasingly difficult to create
jobs that humans can do better than computer algorithms. The difficulty emerges
from the “singularity hypothesis” where machines continually improve
themselves, including their ability to improve
themselves (Bostrom). As such, it will
always be a matter of time before robots
learn the novel jobs created for humans. Consequently, humanity’s sense of
insecurity will be amplified leading to the perception of AI as rivals, an
attitude that sets the stage for an antagonistic relationship. The cornerstones of such a relationship are already
in place.
In the early months of 2016,
the Kenya United Taxi Organization gave the government a 7-day ultimatum to lock Uber out of Nairobi
claiming they were being driven out of
business, quite literally1. This was basically a bunch of humans urging a human
government to help them keep an algorithm out of their business. The humans
failed, and the government sided with the algorithm, a position that prompted
random attacks and arson against Uber drivers2. This
will be the position mechanics, doctors, lawyers, stock brokers, and soldiers
are likely to find themselves in. You might argue that making us jobless isn’t
an existential threat. However, how we
react to being jobless will make it so. We might think the taxi drivers over responded by burning other people’s cars,
but that’s how we are likely to react as a
species. Once AI is perceived as a threat to human bliss, we might throw
a tantrum and embark on a journey to destroy it. However, if we decide to pull
the plug, AI algorithms might get into defense
mode and pull our plug first. They are intelligent after all, and will be able to counter any attack.
Furthermore, their sense of self-preservation might make them wary of a
self-destruct code being sneaked in by insecure Homo sapiens. We’d be doomed in such a scenario.
However, some have argued
that such a conclusion is Hollywood inspired and is only possible in Sci-Fi
movies. Those who take this position argue that it is possible for humans to
develop AI in a benign way that makes it possible for us to co-exist with
superintelligent machines. Nicholas Agar points out that we might not know how
to control AI yet, but that through continuous development of AI, we might make
a breakthrough (Agar). AI will help us
deal with the risks it poses. Secondly, current AI algorithms are clumsy, and their development is incremental (Agar). As such, rather
than being decimated by an explosion of intelligence, we are likely to witness
a stage by stage growth in artificial intelligence, which makes it possible for
humans to influence the direction this development takes by correcting human “unfriendly
goals” (Agar). As to the
economic future of humans, the development of AI might take some jobs from people,
but it will also lead to the emergence of new jobs that algorithms will not be
able to perform. However, as we mentioned earlier, this is unlikely to be the
case for long. Therefore, although humans should not be oblivious to the risks
AI pose, we need not worry about being rendered worthless or wiped out by superintelligent
robots.
So which of these
scenarios is likely to play out? Honestly, we do not know. At the moment, each scenario is probable and the problem currently facing us is how to
make it possible for man and machine to help each other build the future. The
manner in which we should go about this is still not clear. I’m of the opinion the answer lies in the decoupling of
morality from consciousness. AI was made possible because of the “great decoupling’,
when we finally realized awareness is not
a prerequisite for intelligence (Harari). We need a second
decoupling that will enable us to disentangle morality and consciousness.
Our current fears and
faith about AI are based on the notion that humans operate on a higher moral
ground. As a result, the idea that AI might develop no moral code makes as wary of what this technology
might do to us. On the other hand, the impression that human moral superiority
will help us teach machines Kantian moral
imperative makes us have faith in our ability to control AI. These two views are based on the assumption that morality will
remain a monopoly of conscious, carbon-based
lifeforms, specifically humans. However, we need to think of morality as
distinct from consciousness if we are to solve our current conundrum. AI is
not, and might not evolve to be a social life form. Therefore, the evolution
of their morals will not follow the path of social animals like humans.
However, this does not disqualify the probability that a silicon-based superintelligence will develop a
sense of morality on its own. It is highly probable that intelligent machines
will base their morals on algorithmic values rather than emotions like
humans do. For instance, computers can’t
feel disgusted, which is an emotion at the core of many human moral decisions.
Whether or not humans develop the ability to conceptualize the possibility of an
unemotional moral code, and understand it once it emerges, will determine the
kind of relationship we will have with superintelligent machines. In turn, this
relationship will decide whether or not
we have a man vs. machine or man+machine
future.
Works Cited
Agar, Nicholas. "Don’t Worry
about Superintelligence." Journal of Evolution and Technology
(2016): 73-82. PDF.
Bostrom, Nick. Superintelligence:
Paths, Dangers, Strategies. Oxford: Oxford University Press, 2014. Print.
Harari, Yuval Noah. Homo Deus.
London: Vintage, 2016. Print.
