See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/294876416

“The Future of Artificial Intelligence:

Benevolent or Malevolent? Book Reviews of
Michio Kaku, The Future of the Mind: The
Scientifi....

Article in Skeptic · January 2015

CITATIONS READS

0 229

1 author:

George Michael
Westfield State University
89 PUBLICATIONS 140 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

I'm currently working on an edited book on the alt-right. View project

All content following this page was uploaded by George Michael on 18 February 2016.

The user has requested enhancement of the downloaded file.

 RE VIEWS

The Future of
Artificial Intelligence:
Benevolent
or Malevolent?
Reviews of The Future of the Mind:
The Scientific Quest to Understand, Enhance,
and Empower the Mind by Michio Kaku, Doubleday, 2014. Thomas Dunne Books, 2013.
$28.95. 400 pp. $26.99. 335 pp.
and Our Final Invention: Artificial Intelligence SBN-13: 978-0385530828 ISBN: 978-0312622374.

and the End of the Human Era by James Barrat

required increasingly larger brains. The
REVIEWED BY G EORGE MICHAEL development of language was believed
to have accelerated the rise of intelli-
gence insofar as it enhanced abstract
In his 2013 State of the Union recent books examine trends in these thought and the ability to plan and or-
Address, President Barack Obama an- areas of research and their implications. ganize society. With these new capabili-
nounced federal funding for an ambi- In The Future of the Mind Michio ties, humans could join together to form
tious scientific endeavor christened the Kaku, a professor of theoretical physics hunting teams, which increased their
BRAIN (the Brain Research Through Ad- at the City College and City University likelihood of survival and passing on
vancing Innovative Neurotechnologies) of New York, draws upon numerous their genes. The increase in intelligence
Initiative. The $3 billion project seeks to fields, including biotechnology, psy- and expressive capabilities led to the
unlock the secrets of the brain by map- chology, evolutionary theory, robotics, emergence of politics as humans formed
ping its electrical pathways. That same physics, and futurism, to survey what factions to vie for control of the tribe.
year, the European Union unveiled its lies ahead for the human race on the What was essential to this progress was
Human Brain Project, which will use the cusp of what could be a quantum leap the ability to anticipate the future.
world’s largest computers to create a in intelligence. As Kaku explains, the Whereas animals create a model of the
copy of the human brain made of transis- introduction of MRI machines could do world in relation to space and one an-
tors and metal. Generous funding to the for brain research what the telescope other, Kaku develops a “space-time
tune of 1.19 billion euros (about $1.6 bil- did for astronomy. Just as humankind theory of consciousness” for human
lion) has been earmarked for this effort. learned more about the cosmos in the psychology implying that humans, un-
These two ambitious studies could 15 years after the invention of the tele- like other animals, create a model of the
create a windfall by generating new dis- scope than in all of previous history, world in relation to time, both forward
coveries for treating incurable diseases likewise advanced brain scans in the and backward. He argues that humans
and spawning new industries. Concomi- mid-1990s and 2000s have trans- are alone in the animal kingdom in un-
tant with these projects are exciting new formed neuroscience. Physicists played derstanding the concept of tomorrow.
developments in the field of artificial in- an important role in this endeavor as Thus the human brain can be character-
telligence (AI)—that is, computer engi- they were involved in the development ized as an “anticipation machine.”
neering efforts to develop machine-based of a plethora of new diagnostic instru- Kaku employs the metaphor of a
intelligence that can mimic the human ments used for brain scans, including CEO for how the human brain func-
mind. Concrete progress toward this goal magnetic resonance imaging (MRI), tions, in which numerous parties in a
was realized in June of 2014, when it was Electroencephalography (EEG), Com- corporation clamor for the attention of
announced that a computer had just puterized Tomography (CAT), and the the chief executive officer. The notion
passed the “Turing Test”—the ability to Positron Emission Topography (PET). of a singular “I” making all of our deci-
exhibit intelligent behavior indistinguish- Getting to our current level of sions continuously is an illusion cre-
able from that of a human. At a test com- human intelligence involved many evo- ated by our subconscious minds, says
petition organized by Kevin Warwick, a lutionary pathways. Previously in our Kaku; instead, consciousness amounts
so-called “chatterbot” convinced 33 per- evolution, those humans who survived to a maelstrom of events distributed
cent of the judges that it was human and thrived in the grasslands were those throughout our brains. When one com-
with a 13-year old boy’s personality.1 Two who were adept at tool making, which peting process trumps the others, the

volume 20 number 1 2015 WWW.S KEPTI C .COM 57

 R E V IE WS

brain rationalizes the outcome after the less. Duplicating this process in a com- search he interviewed a number of lead-
fact and concocts the impression that a puter is a tall order. In point of fact, the ing scientists in the fields of AI and ro-
single “self” decided the outcome. digital computer is not really a good ana- botics. Although all of his subjects were
Genetic engineering might some- log of the human brain as the latter op- confident that someday all important
day be used to enhance human intelli- erates a highly sophisticated neural decisions governing the lives of humans
gence. By manipulating only a handful network. Unlike a computer, the human would be made by machines, or humans
of genes, it could be possible to increase mind has no fixed architecture; instead, whose intelligence is augmented by ma-
our I.Q. Brain research suggests that a collections of neurons constantly rewire chines, they were uncertain when this
series of genes acting together in com- and reinforce themselves after learning epoch would be reached and what its
plex ways is responsible for the human a task. What is more, we now know implications might be.
intellect. There’s an upper ceiling for today that most human thought actually Much of Barrat’s book is devoted to
how smart we could become based on takes place in the subconscious, which countering the optimism of the so-called
the laws of physics, however, as Kaku still remains somethng of a black box in “singularitarians.” Vernor Vinge first
notes, nature has limited the growth brain research. The conscious part of coined the term singularity in 1993 in an
and development of our brains. For a va- our mind represents only a tiny part of address to NASA called “The Coming
riety of reasons, it is not physically feasi- our computations. Technological Singularity.” The term was
ble to increase human brain size and Kaku asks an important question: then popularized by Ray Kurzweil, a
add to the length of neurons. Thus, he How should we deal with robot con- noted inventor, entrepreneur, and futur-
says, any further enhancement of intelli- sciousness that could decide the future ist who predicted that by the year 2045
gence must come from external means. of the human race? An artificially intelli- we would reach the Singularity—“a fu-
In the field of medicine, brain re- gent entity programmed for self-preserva- ture period during which the pace of
search could increase longevity and en- tion would stop at nothing to prevent technological change will be so rapid, its
hance the quality of life for many someone from pulling the plug. Because impact so deep, that human life will be ir-
patients. Engineers are currently work- of their superior ability to anticipate the reversibly transformed.” As he explained
ing to create a “robo-doc,” which could future, “robots could plot the outcomes in his book, The Singularity is Near, peo-
screen people and give basic medical of many scenarios to find the best way to ple will begin the process of leaving their
advice with 99 percent accuracy almost overthrow humanity.” This ability could biological bodies and melding with com-
for free. Such a device could do much lead the way for a real-life Terminator sce- puters. He predicts that by the end of the
to bring down accelerating healthcare nario. In fact, Predator drones may soon 21st century the non-biological portion of
costs. Through the fusion of robotics be equipped with face recognition tech- our intelligence will be trillions of tril-
and brain research, paralyzed patients nology and permission to fire capabilities lions of times more powerful than un-
could one day use telekinesis to move if it is reasonably confident of the identity aided human intelligence. An unabashed
artificial limbs. Complete exoskeletons of its target. Furthermore, inasmuch as technological optimist, Kurzweil believes
would enable paraplegics to walk about robots are likely to reflect the particular that the singularity will herald a new era
and function like whole people. Taking ethics and moral values of their creators, in human history in which problems
this principle a step further, people Kaku sees the potential for conflict be- such as hunger, disease, and even mortal-
could control androids from pods and tween them, a scenario perhaps not un- ity will be solved. Based on the notion of
live their lives through attractive alter like that depicted in The Transformers accelerating returns, if humans survive
egos in the style of the 2009 movie Sur- movie series. Finally, Kaku speculates on this milestone, the 21st century should
rogates starring Bruce Willis. Perhaps what form advanced extraterrestrial intel- witness technological progress equivalent
AI may even allow people to one day ligence might take. Assuming that once to 200,000 years. Inasmuch as techno-
escape their bodies completely and intelligent life emerges it will continue to logical evolution tends not to occur in
transition to a post-biological existence. advance, then our first contact with supe- linear trends, but rather, exponential
Funding for artificial intelligence rior life outside of Earth could be with in- trends, scientific development will ad-
has gone through cycles of growth and telligent super computer entities that vance so rapidly that the fabric of history
retrenchment. Initial optimism is often have long abandoned their biological will be torn. Singularitarians anticipate a
followed by frustration as scientists real- bodies in exchange for more efficient and future in which AI will allow us to realize
ize the daunting task of reverse-engi- durable computational bodies. our utmost potential.
neering the brain. The two most Whereas Kaku’s tone on AI is The singularitarian movement has
fundamental challenges confronting AI mostly optimistic, James Barrat’s prog- strong religious overtones, which Barrat
are replicating pattern recognition and nosis is dystopian to the point where our argues is overly optimistic. In contrast to
common sense. Our subconscious very existence may be threatened by AI. Kurzweil, Barrat fears that humans will
minds perform trillions of calculations In Our Final Invention, the documentary eventually be left out of this historical
when carrying out pattern recognition filmmaker warns about the looming process and relegated to the dustbin of
exercises, yet the process seems effort- threat of smart machines. For his re- evolution. Holding extreme misgivings

58 SKE PTI C M AG AZ I N E volume 20 number 1 2015

about artificial intelligence, he warns advantage of split-second opportunities pled between 1,000 and 2,000 cen-
that the singularitarians are naïve about in price fluctuations of stocks. In recent trifuges and set Iran’s nuclear weapons
the peril posed by self-aware machines. years, Wall Street has been using agent- program back two years. But as Barrat
The more sanguine scientists believe based financial modeling that simulates warns, malware of this sort does not just
that this process will be friendly and col- the entire stock market, and even the simply go away; thousands of copies of
laborative akin more to a handover than entire economy, to improve forecasting. the virus escaped the Natanz plant and
a takeover; however, Barrat argues that Barrat fears that the intelligence explo- infected other PCs around the world.
such an assumption is misguided. In- sion in the computational finance do- Barrat warns that such cyber operations
stead, he avers that the process will be main will be opaque for at least four are terribly short-sighted and carry a
unpredictable and inscrutable. He fears reasons. First, it will probably take place high risk of blowback. As he explains,
that we could lose control over AI and in various “black box” artificial intelli- now that Stuxnet is out in the public do-
the results could be catastrophic. Hence, gence techniques closed to outsiders. main, it has dramatically lowered the
the ultra-intelligent machine could be Second, the high-bandwidth, millisec- cost of a potential terrorist attack on the
our final invention. ond-fast transmissions will take place U.S. electrical grid to about a million
As Barrat explains, trying to fathom faster than humans can react to them as dollars.
the values of an entity a million times witnessed during the so-called Flash Perhaps in the not-so-distant future,
more intelligent than humans is beyond Crash on May 6 of 2010 when the Dow computers will be autonomous agents
our comprehension. Simply put, the ma- Jones Industrial Average plummeted by making decisions without guidance from
chine will not have human-like motives 1,000 points within minutes. Third, the human programmers. Moreover, the
because it will not have a human psyche. system is extremely complex and thus transition from artificial general intelli-
Though AI may harbor no ill will toward beyond the understanding of most fi- gence to artificial super intelligence
humanity, the latter could get in its way nancial analysts. And finally, any AI sys- could come swiftly and without fore-
and be deemed expendable. He finds it tem implemented on Wall Street would warning, thus we will not have adequate
irrational to assume that an entity far more than likely be treated as propri- time to prepare for it. Once it has access
more intelligent than we are and which etary information and kept secret as to the Internet, an AI entity could find
did not evolve in an ecosystem in which long as it makes money for its creators. the fulfillment of all its needs, not unlike
empathy is rewarded and passed on to In the near future, it is reasonable to as- the scenario depicted in 2014 film Tran-
subsequent generations, will necessarily sume that computer technology will scendence in which Johnny Depp starred
want to protect us. As he argues: have the power to end lives. As Barrat as the mind behind a supercomputer.
points out, semi-autonomous robotic To be safe, Barrat advises that AI
You and I are hundreds of times
drones now kill dozens of people each should be developed with something
smarter than field mice, and share
year on the battlefield. akin to consciousness and human under-
about 90 percent of our DNA with
Nefarious forms of quasi-artificial standing built in. But even this feature
them. But do we consult them before
intelligence already have befallen us. For could be dangerous. After all, a machine
plowing under their dens for agricul-
example, “botnets” that hijack infected could pretend to think like a human and
ture? Do we ask lab monkeys for their
computers (unbeknownst to their users) produce human-like answers it prepared
opinions before we crush their heads
and launch DDOS (distributed denial of to implement its own agenda.
to learn more about sports injuries?
service) attacks are designed to crash Kurzweil has argued that one way to
We don’t hate mice or monkeys, yet we
and/or jam targeted networks. For Bar- limit the potentially dangerous aspects of
treat them cruelly. Superintelligent AI
rat, it would seem to logically follow artificial intelligence is to pair it with hu-
wouldn’t have to hate us to destroy us.
that as AI develops, it will be used for mans through intelligence augmenta-
As Barrat notes, the way we treat our cybercrime. Ominously, cyber-sabotage tion. As AI becomes intimately
closest relatives—the great apes—is not could be directed at critical infrastruc- embedded in our bodies and brains, it
reassuring for those chimpanzees, orang- ture. If, for instance, the power grid will begin to reflect our values. But Bar-
utans, and gorillas that are not already were taken down it would have cata- rat counters that super-intelligence could
bush meat, zoo inmates, or show biz strophic results. As an example of the be a violence multiplier, turning grudges
clowns, the rest are either endangered or great peril posed by semi-autonomous into killings and disagreements into dis-
living on borrowed time. computer programs Barrat cites the case asters, not unlike how a gun can turn a
Even today, computers are respon- of a joint U.S.-Israeli cyber campaign fistfight into murder. Today, much of the
sible for important decisions that affect against Iran dubbed “Olympic Games,” cutting edge AI research is being under-
the economy. In the realm of finance, up which unleashed the Stuxnet computer taken by the Pentagon. The Defense
to 70 percent of Wall Street’s equity virus. Stuxnet was designed to destroy Advanced Research Projects Agency
trades are now made by computerized machinery, specifically the centrifuges (DARPA) has been investigating ways to
high-frequency trading systems—super- in Natanz nuclear enrichment facility in implement artificial intelligence to gain
computers that use algorithms to take Iran. Highly effective, the worm crip- an advantage on the battlefield. Put

volume 20 number 1 2015 WWW.S KEPTI C .COM 59

 R E V IE WS

simply, intelligence augmentation is no

moral fail-safe.

LEAVING TRUTH
Invoking the Precautionary Princi-
ple, Barrat counsels that if the conse-
quences of an action are unknown but
by Keith Sewell judged by some scientists to carry a risk
of being catastrophic, then it is better
not to carry out the action. He concedes,
I think that our most functional criterion however, that relinquishing the pursuit
for identifying proposals as knowledge is of artificial general intelligence is no
physical observation, repeatable on de- longer a viable option. To do otherwise
mand. What I’d like to understand, at would cede the opportunity to rogue na-
last, is the ostensibly more powerful tions and gangsters who might not be as
basis upon which theists seek to deny scrupulous in engineering safeguards
this. They must have one, as their defin- against malevolent AI. There is a deci-
ing proposals stand in direct opposition sive first-mover advantage in AI develop-
to our entire body of on-demand-repeat- ment in the sense that whoever first
able physical observation based knowl- attains it will create the conditions nec-
edge. Reality has no option for showing essary for an intelligence explosion. And
. us, more clearly than it already has, that they can pursue this goal not necessarily
TION the miracles upon which our theists base for malevolent reasons, but because they
I
ED their initial beliefs in their Supernatural will anticipate that their chief competi-
W
NE Beings never really happened. tors, whether corporate or military, will
be doing the same.
To make this challenge explicit, I am not Perhaps the best course of action
merely claiming that the theists are wrong. I’m claiming that they are would be to incrementally integrate
wrong by any criterion through which right and wrong can be coherently components of artificial intelligence
distinguished. This claim is a lot stronger, and it’s testable. For example, with the human brain. The next step in
if Christians can show any functional basis for knowledge-selection that intelligence augmentation would be to
validates the existence and power of Yahweh over his logically exclusive put all of the enhancements contained
alternatives (Allah, Vishnu, Wotan, etc.), or if Muslims can show any in a smart phone inside of us and con-
such basis that preferentially validates Allah, then my claim would be nect it to our brains. A human along
invalidated. with Google is already an example of ar-
tificial super-intelligence. Inasmuch as
Most succinctly, we have never been able to win at the level of “our AI is developed by humans, Kurzweil ar-
truths” against “their truths,” but I think that we can now win at the gues that it will reflect our values. He
level of on-demand-repeatable physical observation vs. our species’ maintains that future machines will still
common-sense concept of “truth” itself. I think that we have had all of be human even if they are not biologi-
the needed philosophical pieces in place, for about the 80 years since cal. To be safe, Barrat recommends ap-
publication of Karl Popper’s Logic of Scientific Discovery, to definitively plying a cluster of defenses that could
call the theist’s bluff at this deepest accessible epistemic level. My mitigate the harmful consequences of
book’s essays therefore argue and provide ammunition for such a bluff malevolent AI, including programming
call, between ourselves and all who still proselytize for emotionally se- in human features, such as ethics and
ductive irrational knowledge systems (systems that can only be propa- emotions. These qualities will probably
gated as “truth”). If I can get enough of you in own my camp to have to be implemented in stages be-
understand and help me to spread this call, then – like Archimedes with cause of the complexity involved, but
his lever – we will start to move the world. by doing so, we could derive enormous
benefits from machine-based intelli-
For more information please visit our website at poppersinversion.org, or buy my gence without being consigned to evolu-
book Leaving Truth as a paperback from Barnes & Noble; or as an eBook from tionary obsolescence.
any of the main e-retailers.
——————————————————————— Refrences
1. Chris Williams, “Ukrainian Teen Created

poppersinversion.org in Lab Passes Turing Test—Famous

Nutty Prof,” The Register, June 9, 2014,
http://www.theregister.co.uk/2014/06
/09/software_passes_turing_test/.

60 SKE PTI C M AG AZ I N E volume 20 number 1 2015

View publication stats