[D66] How to escape feedback

[A..O]

https://mondediplo.com/2018/05/07fake-news

How to escape feedback

By David Napier, mondediplo.com
View Original
May 1st, 2018

The first experiment on the effect of feedback on the social
amplification of messages meant to make Americans react to the threat of
nuclear attack.
by
Graphica Artis · Getty

Winston Churchill once said: ‘History will be kind to me for I intend to
write it.’ That might well be the epitaph for our information age, in
which facts seem to be made as much by the repetition of statements as
by the complex process required to unwind and verify the content of what
we hear. Alan Turing, father of machine learning, might have said as
much, arguing in 1936 that we would reach machine completeness when a
system of instructions given to a machine resulted in it creating an
autonomous duplicate — its own verification.

According to Turing, a computer could be said to ‘think’ if a human
could not distinguish its processing of data from a potential human
response. To do that, it would need an algorithm for learning to sort
random data and its impact on the machine, much as James Maxwell’s
imagined ‘demon’ of 1872 could, by rapidly sorting hot (fast) and cold
(slow) molecules, decrease entropy. If a machine can account for the
disorder of uncertainty, it could be an artificial intelligence — maybe
even a machine intelligence or, better yet, a machine with the ability
to make decisions, like HAL, the murderous computer in Stanley Kubrick’s
2001: A Space Odyssey (1968).

Though the idea that a machine might think is still contested, it has
fascinated scientists, artificial intelligence experts and robotics
innovators for a very long time, because this problem is both
mathematical and physical: if a machine could remake itself, not only
would it overcome disorder, but the information fed into it could result
in an equal or greater output of energy in a system. For a duplicating
machine can multiply a message; and messages can be multiplied endlessly
if machines are built for such tasks.
The stuff of science fiction

While this ability to think may seem fantastical, the problem is more
real than we believe. Thinking machines are not only the stuff of
Kubrick’s science fiction, they are the stuff of real science that can
be uncomfortably stranger than fiction. Professor Kevin Warwick of the
University of Reading has bio-engineered robots that ‘learn’ to refine
tasks as their cells multiply; the ability of a machine to think is
evidenced as the robot’s brain cells grow, making its choices more
systematic and more like behaviour.

But repeating choices by compounding responses organically is quite
different from deciding what to do when choices are unclear. So the
question of whether a machine can think is debatable. The first debate
is about inductive knowledge, what we take to be true based on our
repeated experience of things, and it’s where any pretension of making
artificial life ends. For if a machine could think, we might be the last
to know that it was thinking about us. Thinking independently is proven
not merely by the capacity to create a new set of relations not
otherwise obvious, to use the patent attorney’s criterion for invention,
or to complete a task better than a human can. Independent thought is
also demonstrated by a living being’s ability to suspend and extend its
contemplation and its judgments, and to withhold information, even
openly deceive others, especially those who are not alert and vigilant.

The second debate is over the perceived modernity of the problem of fake
news and what was learned or not from the first experiment that applied
artificial intelligence techniques to a social problem in real time.
That experiment had nothing to do with persuading people to spend money
by intermediating their social relations (Facebook); it did not involve
profiling people to sway their votes by feeding them selected
information (Cambridge Analytica). It was about sensitising Americans to
the dangers of nuclear war.

This first big experiment began in New York in March 1946, when the
Josiah Macy Jr Foundation began convening a group of key figures from
across scientific fields to discuss ‘Circular Causal and Feedback
Mechanisms in Biological and Social Systems’. To the credit of the
foundation, it was among the first ever to bring together diverse
thinkers around emerging problems with significant health and wellbeing
impacts on broad sectors of society. And it was because of this bold
initiative that anthropologists Margaret Mead and Gregory Bateson came
to work with the father of cybernetics, MIT professor and genius Norbert
Wiener, and to meet Claude Shannon, the information and communication
theorist who really should be thought of as the father of the
information age.
‘An ecology of the mind’

Bateson, son of the distinguished English geneticist William Bateson,
had migrated to the US to join the war effort, bringing a keen interest
in group behaviour and psychology after his fieldwork with Margaret Mead
in New Guinea and Indonesia. His research in the US on social
decision-making eventually led to him coining the term ‘double bind’ — a
psychological condition where a person receives conflicting messages,
none with a satisfactory response. Unlike Hobson’s choice — made famous
by a 16th-century Cambridge stable owner who offered only the option to
take it or leave it — the double bind left a person in the anxious state
of having more than one real choice, but all choices being
unsatisfactory. Bateson understood the true test of machine learning.

That’s why the US during the cold war in the 1950s became important for
Bateson; because in times of pervasive social fear there are lots of
problematic choices being made — poor choices that, as Nietzsche had
recognised a hundred years earlier, often meant people choosing bad
meaning over no meaning. For Bateson, there was no good choice possible
in the context of an ever-present sense of anticipation, even social
paranoia. How, he wondered, could people learn to deal with
decision-making in an uncertain world where assessing complex cues and
dealing not only with deception, but with self-deception, was central to
conscious thinking, to self-awareness, and to what he later called an
‘ecology of mind’?

Bateson’s observations were personal for me: not only because I was a
child of the 1950s and had grown up ‘ducking and covering’, the response
that was recommended to nuclear attack, but because years later, near
his death, Bateson wrote to me saying, only partially in jest, that if I
intended to study how religiously minded people in Indonesia received
and responded to messages from the divine through masked rituals, I
ought first to grow a beard. For you never know what messages might come
your way, so it was best to learn a bit first about concealing your
identity when unexpected information led to unanticipated psychological
demands on your sense of self. And that is a lesson, we now must
concede, Facebook has never learned.
Duck and cover: an air raid drill in 1951
Bettmann · Getty

Pioneer though he was, Bateson was not solely responsible for our
understanding of the effects of information and misinformation on public
behaviour in troubling times. His then wife, Margaret Mead, was rapidly
becoming the best-known anthropologist in the US, and already appealed
to many women seeking freedom from domestic oppression; she knew more
about liberated women in other cultures than anyone on earth. Wiener was
no less gifted, and his work on information networks secured his
deserved place as the father of information systems theory.

Wiener interested Mead and Bateson, and they him; for Wiener had not
only honed in on Turing’s ideas about artificial intelligence, but was,
more importantly, concerned with how to measure the social impact of
information. Together the three came to envision the true social power
and potential of the cybernetic feedback loop, the device through which
the effects of a social signal could be recognised and compounded
through machine learning. And it was by working together across their
very broad range of experiences that they realised the social potential
of refining, amplifying and compounding messages inserted recursively as
feedback into an information system.
Simple feedback

All three understood that information meant nothing outside of its
social context. Information needs to resonate socially, as today all
theorists of cybernetics know. In information theory, positive feedback
does not mean getting a good message. It measures only the strength of a
signal as shown in the intensity of feedback it creates and in that
feedback’s ability to refine and compound the effects of a signal.

This means that positive feedback is not an instrument of democracy, but
an amplification device, providing you with more and more versions of
what it evidences you already intuit or suspect to be the case. Like the
queue of videos YouTube lines up ready for you while you watch just one,
the key tasks of feedback are to determine what you are inclined to
follow; feed you experimental variations of that; and, each time,
measure which way your inclinations lead you.

It is an advertising executive’s dream: plain and simple positive
feedback, a thing neither good nor bad. For amplification to occur, it
matters little what the message contains as long as it can be
compounded, and that’s the amoral part; for compounding itself affects
how we humans respond or not to the information put before us.
Cybernetics, it turns out, is just like capitalism, using information,
instead of money, to make more information, instead of or in addition to
more money.

But if the system is amoral, the problems it can engage and amplify are
not; for feedback can push you into more extreme views by catering to
your often unreal sense of self. That’s what makes fake news a big
debate: the real issue is how we are moved by feedback, which takes us
right back to the cold war.

In the US in the 1950s, many citizens thought they could survive a
nuclear attack by hiding in grandma’s root cellar until the fighting
ended, waiting until it all blew over so that robust persons could
emerge and start things anew. Few could imagine the immanent and
catastrophic horror of a nuclear attack or its long-term consequences,
and that was a problem. The US government was afraid of what people did
not know, and they needed to make sure their messages not only made
sense, but could be intensified enough to secure public participation in
what were then believed to be more realistic preparatory measures. Hence
the amplification of nuclear fear and the duck and cover exercises that
made many innocent children paranoid and panic-prone. There is a
credible argument made by sociologist Jackie Orr that modern panic
disorder emerged from duck and cover drills of the 1950s.
The first Universal Turing Machine

That’s where feedback ran away with itself; for along with Bateson, Mead
and Wiener, the 1946 Macy conference also invited John von Neumann,
perhaps the foremost mathematician of his time, who modelled nuclear
explosion devices for the Manhattan Project in Los Alamos, where the
Little Boy and Fat Man bombs that destroyed Hiroshima and Nagasaki in
1945 were developed. He was also a major contributor to quantum
mechanics, game theory and computer modelling; he died without
completing his book The Computer and the Brain, which he felt was much
more important than his nuclear work and which was published
posthumously in 1958. By late 1945, von Neumann had convened a group of
engineers in Princeton to build and programme a digital computer, the
first actual realisation of a Universal Turing Machine. The original
Macy conference organiser, Warren McCulloch, was also instrumental; he
was a neurophysiologist who would focus on cybernetics and the
application of neuroscience to artificial intelligence. He later joined
Wiener at MIT, and his work on automata and network theory is generally
considered as the foundation of modern cybernetic computation.

But unlike the theoreticians, Bateson was emerging from a difficult war
and years of working in remote places on indigenous modes of thought.
With other anthropologists, he had joined the OSS, the clandestine US
organisation, working in the Pacific and Southeast Asia on black
propaganda, including the creation of divisions in enemy society. Though
Bateson had evolved theories on social cohesion and discord, when the
OSS in 1947 became the Central Intelligence Agency (CIA), he had to walk
a line between his political scepticism and his interest in philosophy,
psychology and information systems. His later years in Santa Cruz may
have left him with feelings something like those of Christopher Steele
(of the notorious intelligence report on Donald Trump), another Briton
committed to his best intentions, but alienated by responses to his work
in the US intelligence community.

Then there’s the Swiss contribution: Albert Hoffman, the ‘father’ of
LSD, worked for the Basle laboratories of the pharmaceutical company
Sandoz. It had contracts with the US Food and Drug Administration, which
interested the new CIA: in the paranoid era of cold war, of Joseph
McCarthy, whose threats forced so many to flee the US, the CIA wanted to
know if LSD could make people talk. Its 1953 MKUltra project, in
cooperation with the US Army Biological Warfare Laboratories, licensed
investigators to explore this possibility; the project also supported
the Macy Foundation, headed at the time by former OSS officer Frank
Fremont-Smith.

He later ran the Foundation and became president of the World Federation
of Mental Health; his longstanding interests were in psychosomatic
medicine, and in war-related mental disorders, being well familiar with
the research of the great neurophysiologist, Walter B Cannon, who
discovered the fight-or-flight syndrome. Under Fremont-Smith’s
directorship, by 1959, the Macy Foundation chose to use its former
conference success with feedback to study LSD. The CIA had broad support
for major investigations into the effects of drugs on detainees,
including the use of the psychologically disabled in experiments that
can only be described as cruel.

‘The Many Mothers of Invention’

This is where California comes in. The RAND Corporation had focused from
its beginnings as a thinktank in Santa Monica in 1948 on what the US
might do in a nuclear attack; further north, in Stanford, drug
researchers invited young volunteers to sign up for experiments. Ken
Kesey (author of One Flew Over the Cuckoo’s Nest and patriarch of the
Grateful Dead) was first turned on to LSD by the CIA. ARPA, the US
Defence Department’s Advanced Research Projects Agency, studied
interrogation methods, and had directed funds to determine how to apply
cybernetics to communication networks. It was also responsible for
Arpanet, the ‘nuclear-bomb aftermath’ inter-university precursor to the
Internet. Such early public-private partnerships were fast growing in
postwar California. RAND set itself up in Santa Monica, and eventually
among other places in the Bay area, Pittsburgh and Boston, homes to
Stanford, Carnegie Mellon and MIT, the big university players in
artificial intelligence, computer engineering, robotics and machine
learning.

But the connection between computer modelling, artificial intelligence
and drugs was about more than seeing how drugs affected behaviour in
experimental settings useful for military intelligence. In 1967 Douglas
Engelbart, father of computer-human interactions and funded by ARPA,
applied for a patent for the computer mouse; he had also, while on LSD,
envisioned computer bootstrapping (the ability to self-start processes).
New technology and psychedelia met the economy of 1960s California,
where self-oriented transcendentalists joined hands with entrepreneurial
capitalists. And that’s how we got Steve Jobs, mindful meditation in the
workplace, iPhones, and all the rest.

Shakespeare wrote: ‘Misery brings a man strange bedfellows.’ The
convergence of new left and old right on the issue of self-centred
individual freedom through new-age technology led to an ideology of
liberation from government and an odd marriage between the new left and
extremism. Any home schooling website shows a stronghold of independent,
libertarian strange bedfellows: rightwing militants, the Steve Bannons
of the world, and the most liberal people, including some Bernie Sanders
anti-establishmenters who voted for Trump when Sanders lost the
primaries. I know their sentiments well, having served in Sanders’s
thinktank when he moved from being mayor of Burlington, Vermont, into
national politics. Little divides the radical left and the radical right
in US anti-government sentiment. That’s because their mutual anger at
the spineless impotence of politicians controlled by special interests
‘trumps’ their freedom divide over welfare and autonomy.

By the dawn of the 1960s, any prospect of turning the clock back on
information and misinformation had already long disappeared. While the
US government saw the potential of cybernetics for sensitising a naïve
public about the real threat of nuclear conflict, it also realised that
measuring and nourishing feedback could help US intelligence, as it now
does Russian intelligence, to foment social unrest abroad. It was the
very amorality of feedback that made it reliable and for so many years
gave artificial intelligence a bad name in humanitarian circles.
Amplifying social responses through what used to be called propaganda
was not a nice thing to do to people; especially because information can
be compounded and exaggerated endlessly. This seems to have been lost on
Silicon Valley, which oddly to this day still thinks of open access as a
new form of democracy.

It is not those ignorant of history who are doomed to repeat it, but
machines that know your history all too well

It isn’t. Had contemporary Silicon Valley pioneers been less enamoured
of their Teslas, t-shirts, and global money making, and more focused on
local history, they could have learned a lot from the University of
California archives where 99 boxes of Bateson’s Santa Cruz papers are
kept. Cybernetics has taught us that history will amplify itself
magnificently, and divide us if we remain naïve about its ability to
repeat the bad along with the good. It is not those ignorant of history
who are doomed to repeat it, but machines that know your history all too
well.

Information is not innocent

That’s where our current problem with social media lies. Since
information is never value-free, we should not consider it innocent or
remain passive to its presence. There are machines already out there
that are much smarter than idle, complacent humans who latch on to
clichés about politics or weaponry that might let them plan to defend
themselves and their autonomy. The independence inherent in open access,
which can result in anything true or false being fed out in search of
feedback, far outpaces in social impact more traditional concepts of
freedom. This leads to a simple realisation: idleness allowed so many to
do nothing while 270m fake Facebook accounts worked to amplify
prejudice. The most conservative, supposedly free, people in the US were
led to serve the foreign interests they historically most feared.

Yes, there is more here than a message about self-deception. The
recursive social feedback loops first imagined by Wiener, Bateson and
Mead have now turned the idle and ignorant into the managers of their
own undoing. While alert humans are still far more advanced than
machines, those who passively sign on for social media have already
given up the argument; for when people relinquish their autonomy, they
become vulnerable to colluding in their own deception. This is
especially true in an age of information where machines position
themselves between those who are capable of thinking critically and
those who aren’t.

History right now needs critical attention, for if we blindly accept a
comfortable view of history, it is we who are doomed to repeat it. The
most constructive way to respond to what social media have done with
your data, and done to you as a person, is to find a dozen people you
have not already submitted to feedback loops when you linked them online
as friends, and invite these new and real acquaintances all by snail
mail to a warm, hopeful dinner in your home.

Courage, friends: there is so much of life to enjoy offline. Meeting
people, not virtual friends, can open more windows than any computer
screen or altered state of consciousness, especially when the room needs
fresh air.