Robot wars: New film explores limits of AI...
Alex Garland's film Ex Machina explores the limits of
artificial intelligence - but how close are we to machines outsmarting man?
Having watched the new film, Rhodri Marsden found himself
celebrating the joys of humanity
By Rhodri Marsden
Author Biography
Thursday 22 January 2015
My girlfriend casually informed me, apropos of nothing,
that she'd washed her hair using shower gel rather than shampoo. I raised an
eyebrow. "Well, I guess it's all the same," I said. "It's all
detergent, isn't it? Shampoo, shower gel…" "Cillit Bang," she
added, prompting 30 seconds of sporadic giggling at the very idea. At that
moment, I could be pretty certain that neither of us were robots. Such
exchanges feel like the essence of being human: unpredictable behaviour, saying
things that don't need saying, reaching absurd conclusions and experiencing joy
as a consequence. Good times.
In the film Ex Machina, released this week, a young coder
finds himself in the company of an attractive robotic female entity named Ava,
built by his boss. The film's narrative hangs around the apparent
self-awareness possessed by Ava; at times, it's evident that she passes the
Turing Test, the term given to the various ways in which artificial
intelligence (AI) is measured against that of humans.
But is Ava actually experiencing consciousness? Can she
feel? Is she capable of making light-hearted quips about bathroom cleaner and
enjoying the reaction said quip prompts in others? Many scientists doubt, even
given the extraordinary advances in technology awaiting us, that AI can
realistically match the combined intelligence and consciousness of human
beings. Others believe fervently that it can, and that it poses profound
questions for the future of humanity. That moment of attainment is known as The
Singularity.
One of the inherent difficulties of debating The
Singularity is that humans are incapable of predicting the nature of an
intellect that will exceed our own. We have no real idea of what The
Singularity means for us. That doesn't, of course, stop scientists, radical
thinkers, philosophers and AI experts making best guesses, and their visions
range from the utopian to the hellish.
Ray Kurzweil, Google's director of engineering and the
man perhaps most closely associated with The Singularity term, is almost
evangelical in his optimism. He excitedly outlines a vision of the future where
man and machine combine, extending our lifespans indefinitely and making
optimum use of everything around us. Science fiction writers and film-makers
have, over the years, sounded a more cautious note in the interest of dramatic
tension.
Some scientists, meanwhile, warn that any entity with
greater intelligence than ours is unlikely to have our best interests at heart.
"If we're lucky, they'll treat us as pets," says Dr. Paul Saffo of
Stanford University in The Singularity, a documentary film by Doug Wolens.
"If we're very unlucky, they'll treat us as food."
The speed at which The Singularity is approaching is,
according to Kurzweil, driven by the nature of the growth in technological
power – "exponentials" – in relation to our own relatively static
intelligence. For example, Moore's Law, formulated in 1965 by the co-founder of
Intel, Gordon Moore, broadly states that computer speed doubles every two
years, and that's still holding true.
Some argue that the overall rate of growth is slowing,
some say it's speeding up, but there's enough confidence knocking about for a
few dates to have been assigned to The Singularity.
Kurzweil pegs it to the year 2045, by which time he
believes that one computer will possess equivalent intelligence to every human
being on the planet combined. These kinds of comparisons, however, rely on the
assumption that the brain can be represented in terms of computing hardware – a
philosophy known as Strong AI.
Advocates say there's nothing intrinsically special about
living matter that prevents it from being modelled; the brain, after all,
processes inputs, delivers outputs and possesses memory – though, yes, with
somewhere approaching 100 billion neurons in a human brain and more possible
connections than there are atoms in the universe, that would be a formidable
computational task. However, that task is made even more complex by those
thorny questions of sentience, self-awareness and consciousness, described by
Sir Nigel Shadbolt, professor of artificial intelligence at the University of
Southampton, as the "Hard Problem".
"We have no idea how much sentience has to do with
being ensnared in the body we have," he says. "We can't even come to
an agreement about where, in the evolution of actual organic life, anything
like self-awareness kicks in. We're building super-smart micro-intelligences –
we call them AI – that can recognise your face, translate from one language to
another, learn to play games and get better at them. But we have no clue about
what a general theory of intelligence is, let alone self-awareness. Why, come
2045, that should that suddenly change... I don't think that matching the
processing power of the computer to the brain is going to give us any better
insight."
Mark Bishop, professor of cognitive computing at
Goldsmiths, is on the same side. "If you accept the premise that
everything about the brain is explainable by a computer program," he says,
"then Kurzweil is absolutely right. But I believe that there are many
strong grounds for doubting that claim. No matter how good the AI, the computer
will never genuinely understand. It might appear to, but only in the sense that
a small child might appear to understand an adult's joke when it laughs
appropriately at a dinner party."
Bishop outlines three arguments that address the question
of consciousness and computing. The first, by John Searle, dates from 1980 and
is known as the Chinese Room; if a computer convinces a Chinese speaker that it
understands Chinese by responding perfectly to their questions, it has passed
the Turing Test. But does it really understand Chinese, or does it only
simulate understanding? The second is Bishop's own argument from his 2002
paper, Dancing With Pixies. "If it's the case that an execution of a
computer program instantiates what it feels like to be human," he says,
"experiencing pain, smelling the beautiful perfume of a long-lost lover –
then phenomenal consciousness must be everywhere. In a cup of tea, in the chair
you're sitting on."
This philosophical position – known as
"panpsychism" – that all physical entities have mental attributes, is
one that Bishop sees as Strong AI's absurd conclusion. Shadbolt agrees.
"Exponentials have delivered remarkable capability," he says,
"but none of that remarkable capability is sitting there reflecting on
what very dull creatures we are. Not even slightly."
The third argument Bishop makes is that there's something
about human creativity that computers just don't get. While a computer program
can compose new scores in the style of JS Bach, that sound plausibly like Bach
compositions, it doesn't design a whole new style of composition. "It
might create paintings in the style of Monet," he says, "but it
couldn't come up with, say, Duchamp's urinal. It isn't clear to me at all where
that degree of computational creativity can come from."
By contrast, one of the scientific advisers for Ex
Machina, geneticist Adam Rutherford, describes the consciousness of Ava's
character as being a mere "couple of conceptual breakthroughs" away.
So it's evident that there's profound disagreement and some deeply entrenched opinions
about the likelihood, let alone the timing, of The Singularity. Kurzweil
rejects the views of many of his critics, saying that they can't comprehend the
exponential rate of technological change, and those who subscribe to his views
dismiss counter-arguments as being weighed down by religious baggage. However,
ironically, Kurzweil's vision of eternal life (where realigning misaligned
molecules can prevent disease, and where we can be physically augmented by
technology) almost seems to be driven by a semi-religious instinct –
specifically, a fear of death.
"People who reject religion are often very scared of
death," says Bishop. "It seems there's a deep lack, and some people
replace that by buying uncritically into a computational dream, a dream of robot
minds and silicon heaven."
Bishop also stresses that there are emergent ways of
looking at the problem of consciousness that bypass that dream. "The
dominant metaphor for the mind over the past 50 years has been the computer,
but there are many exciting new avenues where scientists are looking at the way
consciousness arises without buying into this computational metaphor – which,
in my view, has had its time."
The Singularity and its consequences are almost
inevitably addressed in the media with slightly hysterical overtones. When the
co-founder of Paypal, Elon Musk, recently donated $10m (£6.6m) towards research
into the safety of AI, it was described in one newspaper as a move to
"prevent a robot uprising". And while a post-Singularity world is by
its very nature unfathomable, Shadbolt stresses the need for care, even among
those who don't believe in the concept.
"There are lots of ways of being smart that aren't
smart like us," he says. "A robot doesn't have to have self-awareness
to be autonomous and capable of creating havoc. It's right to come out and say,
for example, that you shouldn't build a self-replicating, single-mission
system." Philosopher Nick Bostrom outlined one such system: a
superintelligence whose goal is the manufacture of paperclips. Let loose, it
would have no reason not to transform the solar system into a giant paperclip
manufacturing facility.
"And I'd suggest," laughs Shadbolt, "that
Earth would be a duller place as a result."
The arguments surrounding The Singularity are deeply
compelling, not least because they ask us to assess what it means to be human.
In a landmark essay for Wired magazine in 2000, Bill Joy, co-founder of Sun
Microsystems, wrote a heartfelt and somewhat anxious response to Kurzweil's
vision, during which he recalled the penultimate scene in Woody Allen's film,
Manhattan. Allen considers why life is worth living: for him, it's Groucho
Marx, the second movement of the Jupiter Symphony, Swedish movies, the crabs at
Sam Wo's and so on.
"Each of us has our precious things," wrote
Joy, "and as we care for them we locate the essence of our humanity."
I briefly glimpsed my own humanity in a joke about
washing one's hair using Cillit Bang – and unlike The Singularity, it was a
very small thing. But maybe it's the small things, not the big ones, that are
most important.
'Ex Machina' is on general release from today
Comments
Post a Comment