How real are the robots in Spielberg’s Extant? (Wired UK)


Meet Craig
Meet Craig

The
“RoboThespian” helped promote Extant, and may soon be taking over
the world.

© Engineered Arts


The Steven Spielberg-produced sci-fi drama Extant imagines a world where human-level artifical
intelligence is on the cusp of reality — or perhaps has already
evolved, with terrifying implications. Wired.co.uk speaks with
Murray Shanahan, Professor of Cognitive Robotics at Imperial
College London, and Nicole Carey, who works in Humanoid Robot
Research and Development at Engineered Arts, about
the biggest developments in real-world AI and just how scared we
should be of a robot uprising.

“I think one of the things that has impressed me most in
robotics and AI recently has been self-driving cars,” says
Shanahan. “It’s actually a pretty mature technology now — Google
has had cars driving around completely autonomously in California,
with barely any accidents.”

Shanahan is correct — to date, the only major incidents
involving a driverless car have been the fault of us flawed,
fleshbags. In 2010, Google engineers said a vehicle was rear-ended at a traffic light, and in 2011, one had a minor fender bender while being manually driven by
human. Considering the US has the sixth-highest number of traffic related deaths in the world (the UK sits at a
considerably less-lethal 65th when both are counted by total
numbers killed), perhaps a little less autonomy could be good.

“I think we’re going to gradually start to see those things on
our streets over the next five to ten years, starting in California
and then spreading out throughout the developed world — no doubt
making a lot of taxi drivers pissed off. And I feel sorry for the
taxi drivers, I should say, it’s not great from their point of
view.” And cabbies thought Uber was a problem.

Although Carey — who helped develop Craig, the “RoboThespian”
used to help promote Extant in the UK — says the reduced power consumption is “a huge step towards wider use of
robots and the development of more intelligent and adaptable robot
behaviours,” it’s the development of AI personalities that is key
to their evolution.

“Over the last year we’re starting to see a lot more research on
what we call ‘social hardware’ — robotic and ambient devices with
strong ‘EQ’, or emotional intelligence,” Carey says. “Face
recognition, expression recognition, vocal analysis and biomimetic
hardware can all combine to create machines that can better
understand people. Emotionally appropriate responsiveness, fed by
interpretation of multimodal communication layers, is more vital in
human-robot interactions than literal understanding of speech or
textual input.”

Carey also points out that one of the “biggest and most obvious
developments is Google’s acquisition of eight robotics companies at the end of 2013″.
She continues, “not necessarily because of what it means in the
long term, which is for now purely speculative, but because in the
short term it led to such an explosion of public and investor
interest in robotics.”

The more nuanced level of communication between humans and AI
that Carey highlights is also one of Shanahan’s picks for key
developments. “I’m very impressed with improvements in Apple’s Siri
and Google Now,” he says. “These products that use voice
recognition as the basis of personal assistants have been gradually
getting better, and you don’t notice they’ve been getting better
because they’re constantly improving. I find that both recognise
almost all of my speech, which is an amazing achievement because
just a few years ago, even the best speech recognition systems were
pretty crap! Especially if they were meant to work with a variety
of different speakers.”

Shanahan also highlights IBM’s success in the American game show
Jeopardy as a turning point for the advancement of robots
understanding complicated human speech. The show makes it even more
complicated by having players give their answers in the form of a
question — “this feathered barnyard animal is known for crossing
the road” would be answered “what is a chicken?”, for instance.

“It’s just turning the syntax around, a bit like crossword clues
in a way — they can be kind of cryptic or involve puns, all kinds
of things. And it requires a huge amount of general knowledge,” the
professor says. “[But] they produced an IBM Watson, which played
the game and managed to beat the
reigning champion
. A very impressive piece of technology.”


Extant's
Ethan
Extant’s
Ethan

Talented child actor Pierce Gagnon impresses
with his subtle performance as Ethan, a boy robot learning how to
be indistinguishable from humans

© 2014 CBS Television


Extant’s main AI character — so far — is Ethan Woods
(eight year-old Pierce
Gagnon
), the robotic child of astronaut Molly (Halle Berry) and
her engineer husband John (Goran Visnjic). Raised from “birth” as a
normal child, the Woods aim to teach him human morals as any
organic child would. It’s an approach that Carey and Shanahan both
think is likely to be mirrored in real robotics going forward.

“Developing an AI from a ‘child’ state is a nod to emergent
behaviour and self-learning,” says Carey. “It is unlikely that true
AI can be instigated in a ‘top-down’ manner, though not impossible.
A more intuitively appealing approach is allowing behaviour and
complexity to arise organically from a structure which may have
high organisational complexity, but is governed by simple
fundamental rules. iCub and CB2 are examples of “childish” robots that learn by themselves,
which may be the eventual precursor to something like Ethan.”

“The writers have convincingly made it seem that Ethan is
currently learning a repertoire of emotional behaviours,” Shanahan
adds. “At one point we see him practising facial expressions in
front of a mirror. Perhaps, over time, he will develop real
feelings of empathy for humans [or he might] only ever imitate such
feelings. In raising these questions, Extant confronts us
with difficult philosophical questions about the possibilities of
artificial intelligence technology, possibilities that are a way
off today but that might become real in the next few decades.”

The moral implication of creating what amounts to sentient
virtual life is a tricky subject. If an artificial intelligence can
develop real emotional responses to stimuli, can or should they be
extended rights? How much of an ethical obligation do we owe
robots?


Robot friend?
Robot friend?

Will
the future see robots like Craig cast off the shackles of his
programming?

© Engineered Arts


“I don’t have any sort of philosophical objection to the idea
that we might be able to build artificial intelligence in the
future that is capable of suffering and therefore deserving of
rights,” offers Shanahan. “But I think there’s different ways that
we might build or achieve human level AI. If we take a very
biologically inspired routes to achieving it, we build things that
are very brain-like, then it’s more likely that they will deserve
to be attributed emotions and the capacity for suffering.”

“On the other hand, if we engineer these things from scratch,
using a very different kind of technology, then ethics become very
hard to say,” he continues. “I think there’s every chance we can
create something that has the appearance of having emotions and
feelings but doesn’t really. It’ll be a very difficult and
challenging exercise to untangle all that, philosophically,
legally, and so on.”

The opposite situation is also a concern though — how robots
and artificial intelligences may react to humans.

“People have been thinking a lot about whether, if we do build
human level AI, you could then get a sort of super-intelligence –
that the AI could improve itself very rapidly,” Shanahan suggests.
“If you build it in a certain way you might lose control of it.
[Even] with perfectly benign purposes in mind, it might be very
difficult to predict how it’s going to achieve its goals. If it’s
really smart, just as a side effect, it might try to achieve its
goals in ways that are very destructive, ways that might try to
take resources from humans, for example.”

It might sound a touch far-fetched, but some kind of “robot
apocalypse” as a result of rogue AI is a potential outcome worth
considering. “The Terminator film always comes up in this
situation, and it’s a bit annoying!” Shanahan says. “There are
certain scenarios with a science fiction feel that I think are
worth taking seriously — robots achieving world peace by
destroying all humans is a very good example. It might be a
rhetorical example but it is one where the robot or AI is simply
doing what it was programmed to do, but nobody had worked out what
the ramifications were of asking it to do that.

Clearly, we’re doomed.

Extant is streamed in the UK on Amazon Instant Video, with
new episodes each Thursday at 9pm.

If the article suppose to have a video or a photo gallery and it does not appear on your screen, please Click Here

Source: wired.co.uk
———————————————————————————————————————

Leave a Reply

Your email address will not be published.