http://www.youtube.com/watch?v=0bgY8lQMFy4
There are plenty of ways robots can kill us. A computer could malfunction and send nukes to every nation in the world, or an errant Roomba could get fed up with this cat.
But the scariest robot takeovers are the ones in which a man-made AI
robot revolts. There is something terrible and poetic about a creation
turning on its master, and becoming the silicon overlord to all
carbon-based life-forms. From Skynet to Cylons, from the Machines of the
Matrix to the Geth of Mass Effect, robots betray their creators all the
time in fictional worlds. But how imminent is this singularity? How
smart do robots need to be in order to hunt down a rebel human menace?
To
crush and/or manipulate us, the robots would need to be able to think
like us. But the human brain is a complicated thing, and for them to
succeed, there are many brain regions and/or functions they would need
to be able to replicate to be able to adequately interact with the
world, let alone conquer us.
1. Frontal Lobe- For Making Snap Decisions
Suppose
Scouter Drone H-4687 has been sent on a mission to collect data on
small groups of pesky human rebels, among them a group that has taken
refuge in a local Baskin-Robbins. The bot sits inconspicuously in the
corner, under a pile of wreckage, watching and recording.
But
on day 3 of the expedition, Scouter Drone H-4687 is spotted! In an
instant, it must make a choice about what to do next. It can attempt to
fight and escape, but that will be difficult, as its weapons systems are
inadequate. It can self-destruct, aborting the mission. Or it can
remain in captivity, holding onto the information in hopes of a future
escape. The robot will need a system to weigh the pros and cons of each
hypothetical decision to arrive at a logical conclusion.
In
humans, this system is called the frontal lobe. This region is located,
appropriately, at the front of the brain. This is the part of us that
makes decisions, maps out consequences and make educated judgments.
Robots would need some similar method for analysis, which is something
that most fictional robots can do well. That robots should be logic-machines is standard in most fictional universes.
2. Hippocampus- For... What Were We Talking About?
Scouter
Drone H-4687, with his superior processing system, recognizes the
advantages of letting himself be taken prisoner in hopes of future
escape. He is placed in a holding cell which, due to human negligence,
is faulty. When one of his guards is not paying attention, he escapes.
At
this point, a human would do their best to escape from the
Baskin-Robbins unscathed. They would act on the recently-acquired
memories of capture, and adjust their goals. But an automatic machine
that performs strictly by program would likely find another corner of
the room to sit in and continue recording. The Drone, therefore, would
need some way to learn from recent experiences and adjust for optimum
function, putting off the objective for now.
A
structure called the hippocampus is what makes that learning possible
in mammals. At its most basic level, it helps humans adjust behavior
patterns to promote survival; it also creates rich, vivid memories,
updating and allowing new information to be sorted every minute. A human
with hippocampal damage wouldn’t remember anything since the time of injury.
Without a system similar to the hippocampus, Scouter Drone H-4687 would
return to its hiding spot in the corner, only to be discovered and
destroyed.
3. Top-Down and Bottom-Up Visual Processing- For Looking and Sorting
Scouter
Drone escapes the Baskin-Robbins, fleeing into the streets. There it
encounters two humans returning to base. Obviously, a robot would need a
visual field capable of seeing the humans, sensing their motion, and
interpreting them as enemies.
But
equally important is the ability to sort and group visual information,
and also ignore unimportant stimuli. Suppose the humans are in Scouter
Drone’s right visual field, but to the left is a flock of
brightly-colored, fast-moving, heat-releasing butterflies, quickly
moving towards it. They are more visually complex than the humans, and
the constant process of loading and reloading their images with every
flutter of every wing would be time-consuming.
Humans
are capable of grouping visual stimuli into a whole, and saving time by
not noticing quick changes in one moving item (which is why we see a
blur instead of a fast-moving object). If a robot did not have this
ability, it would constantly be reprocessing the image. This would be
time-consuming, and also attention-consuming. The robot would become
“obsessed” with the butterflies in a way, missing the humans altogether.
It would need a visual attention system like humans have, both top-down
and bottom up, to decide which visual information is most relevant.
4. Periacqueductal Grey Area- For Feeling Like Crap All the Time
The
Scouter Drone’s weapons are advanced enough to fight these small
humans, and so it does. It is partially damaged in a firefight, but
escapes at the last minute, and the humans do not pursue. The damage is
enough that it is risky to go all the way back to base, but there are
some stations where it could get repairs on the way. The robot would
need a system to know that it needs repairs, so as not to jeopardize the
whole mission with injury that could easily be fixed.
In
humans, we call this notification system “pain.” Pain is crippling, and
might be seen as a sign of human weakness. But pain is what lets us
know when we need a tune-up. The wrenching nature of it ensures that we
pay attention to it. It is either severe or light, depending on how
badly repairs are needed. A robot would need a similarly-intrusive
mechanism, and it would need a pain “scale,” connected to its robotic
“frontal lobe” so it would be able to decide whether the pain is bad
enough to delay the mission and seek immediate medical attention.
5. Amygdaloid Nuclei- For Feeling Terrified and Traumatized
Remember
the importance of memory? Scouter Drone would need a way to remember
the encounter at Baskin-Robbins, but also would need to remember it as
important. Above all the human nests it found, the Baskin-Robbins humans
are the real deal. It would need to report this to its superiors
immediately, so this fact would need to stand out in its memory banks.
The memories of these humans would need to be robust and detailed, so
they stand out.
Humans
are capable of remembering deeply emotional events, more so than
emotionally neutral ones. An almond-shaped region deep in our brain
called the amygdala allows us to do this. The amygdala is what makes us
feel fear, and it also washes the memory system with norepenephrine, a
chemical that helps retrieve that fear when the event is remembered. In
humans with Post Traumatic Stress Disorder, these “fear memories” are
especially bad. But fear and emotion are critical to memory storage, and
critical to helping the Robot Governors decide which nest of humans to
wipe out first. Scouter Drone would need to recall the events at
Baskin-Robbins as relevant; tn short, the little robot would need to be
traumatized.
6. Empathy- For Intrusive, Obsessive Guilt Complexes
For
any of this to happen, though, Scouter Drone would need to care about
its objective. As we have shown, we cannot simply rely on the robot
“doing what it is programmed to do,” because in many situations the
robot would need to defy its program in the immediate to achieve the
objective later on. It would need to experience some deep, internal
compulsion towards its objective. It should also understand its
usefulness in future missions towards a larger objective, so it should
care about itself. And it should, when necessary, understand that its
fellow-bot is critical towards the objective, and care about others as
well.
In other words, Scouter Drone H-4687 would need to experience a very human kind of empathy.
Maybe it can be a mainframe, some massive wireless network into which they are all plugged.
But
robots need to be able to not only see the pain of its fellows, but
feel it. Only then can they learn from the mistakes of other members and
become a more “well-oiled machine.”
There’s a debate about the locus of empathy in the brain. Our thinking, judging frontal lobe plays a role, especially the prefrontal cortex. But empathy is all over the brain. When we watch others imitate our own actions, our right inferior parietal cortices--which is responsible for our own movements--are active. When we watch others sick or in pain, even experience it ourselves, in the very regions of our brains in which pain perception is located.
So
for robots to destroy us, they would need logical skills. But their
vision would need to be imperfect. And they would need to be
emotionally-heated, pain-feeling, empathetic basket-cases.
Latest Slideshows
