18 September 2014
NOTE: This article was
posted one day later than scheduled due to several security issues that made an
earlier posting impossible. I am sorry
for the delay and appreciate the patience of those who regularly read and/or
follow this weblog. This particular blog
has had more than its share of security issues.
I am a bit puzzled by the attention.
HOPE AND FEAR
Technologists
at Harvard are well on their way to developing the first robotic bee. They envision a robotic bee with all the abilities
of the organic original: the honeybee.
Begun in
2009, Harvard’s “Micro Air Vehicles Project” is now using titanium and plastic
to replicate the functions, if not the appearance, of the familiar
honeybee. The robo-bee pops up, complete
with wings, from a quarter-sized metal disk.
The developers predict that these “bees” will be engineered to fly in
swarms, live in artificial hives, and coordinate both their target locations
and pollination methodologies. [2]
In fact,
this program’s goals, if achieved, would produce swarms of robotic bees of such
organization and efficiency that one writer expressed the wish that the project
spokesperson add the phrase “for the good of all mankind” to each progress
report. Without it, readers might be
reminded of all the movies “about technology that eventually destroys mankind.”
While
science fiction films have suggested the replacement of human beings with
robots, films have never “warned” us about the sinister side of the robo-bee. Imagine a robotic “Stepford Bee” hiding
quietly in the wings plotting an unfortunate end for the last of the world’s honey
bees. And, then, a “brave new”
technological world without any “real” honey bees at all!
There is
something definitely creepy about humanly engineered mechanical bees pollinating
crops grown from humanly engineered seeds.
One writer described the disturbing vision as “swarms of tiny robot bees
. . . pollinating those vast dystopian fields of GMO cash crops.”
NOT SO FAST
To read some articles, this robotic bee has
not only been perfected, but is poised the replace its natural counterpart in a
brave new world full of disconcerting, mechanical replicas of the familiar and
comfortable wildlife around us. However,
that future is definitely . . . in the future. [1]
In the
1950’s, futurists predicted that we would all be operating flying automobiles
by 1970. Similarly, the prediction of
working robotic honeybees may be an optimistic fantasy. But if the goal is never reached, it will be
for no lack of effort on the part of the Harvard researchers. But there are many hurdles, challenges, and
obstacles.
With robotic insects, flight itself is the
biggest challenge. While bird-sized
flying drones are being perfected with relative success, flying insect ‘bots
present a special aerodynamic problem.
It’s the
size.
If you
shrink a bird-sized drone down to the size of an insect -- it won’t fly. A roboticist at the University of California
at Berkeley, Ronald Fearing, told the Washington
Post that “the rules of aerodynamics change” with an object as small as an
insect. [5]
Unlike bird wings, insect-sized wings
must move with amazing precision.
Replicating these precise wing movements is a formidable engineering
challenge. In fact, scientists only
recently came to understand how insects fly at all. Compounding these problems, such precision
wing movements require yet larger supplies of portable power. [6]
In order to
create a robot that does what a honeybee does, the ‘bot must be the same size
as a honeybee. Currently, no
lightweight, portable power source exists with both the small size and large energy
output needed by the robo-bee. But even
with a suitable power source, the ‘bot must also be equipped with a portable
guidance system. And there is no
guidance system small enough, and lightweight enough, to do the job. [3]
For now, Robo-Bee
is a sensation because it can fly. But the word “fly” is used in the most
restricted and technical sense. For most of the last few years, Robo-Bee
has been able to flap its wings, and rise into the air – “fly.”
However, when it does, it shoots from its starting position across
the room and crashes into the nearest wall. Flight over. Total
flight time – about a second.
This
constant crashing is even more discouraging when you realize that the current
prototype is stabilized by a fixed wire. And, without that practical,
portable power supply, Robo-Bee still needs “a power cord.” Figuratively
speaking, you still have to plug it into the wall.
Recently, however, researchers have
figured out how to guide the robo-bee in flight. Now, with the latest
guidance breakthrough, the robo-bee can be made “to pitch and roll in a
predetermined direction” and, then, it crashes into the nearest wall.
Discouraged yet? Well, to their credit, the would-be developers of Robo-Bee aren’t the least bit discouraged. And, as modest as the current Robo-Bee’s performance may be, it’s an incredible achievement. Only with the persistence of the project’s engineers have a host of seemingly impossible challenges been met and problems resolved.
Discouraged yet? Well, to their credit, the would-be developers of Robo-Bee aren’t the least bit discouraged. And, as modest as the current Robo-Bee’s performance may be, it’s an incredible achievement. Only with the persistence of the project’s engineers have a host of seemingly impossible challenges been met and problems resolved.
Progress
has, and will, be made through a series of small advances over a long period of
time. So, the rumored release of a swarm
of robotic bees to replace our honeybees is far, far away. [4]
It will be a long time before the first Robo-Bee rolls off the assembly
line, flies into the fields, and begins pollinating.
ANOTHER HORIZON
Whether a
particular article expresses hopeful optimism or fearful apprehension about a
“future” robotic bee, I seem to hear the same description – again and
again. The list of goals extends to the
ability to fly in swarms and coordinate swarm maneuvers and strategy. But a robotic honeybee needs more. Much more.
A HONEYBEE’S BRAIN
After all
of the above issues are resolved and all the goals reached, there will still be
something missing.
Even
with an on-board computer to direct its flight, how will the Robo-Bee pollinate
flowers? Just think about it. To do so, these robots would have to
see and smell. They’d have to master the varied challenges of the
pollination of each individual bloom. To do that, these ‘bots would
almost have to be able to . . . think.
How are they going to do that? Well, the members of the Green Brain
Project are glad you asked them that question.
Researchers
in Great Britain, specifically, at the Universities of Sheffield and Sussex not
only know the question but, about a year ago, decided to do something about answering
it. In an article describing the
project, George
Dvorsky, reports that, late last year, the Engineering
and Physical Sciences Research Council (EPSRC) put up £1 million (USD
$1,614,700) for the development and creation of the “first
accurate computer simulation of a honey bee brain.”
But, when
you consider the challenging goal of the project, even this “front money” is
not so very much. After all, the project
couldn’t afford the kind of computer muscle that would seem to be needed to
tackle a job like this. However, a creative solution to the computer
problem has been provided by California’s NVIDIA. That corporation
will provide the project with a number of high-performance graphical processing
units called GPU accelerators. This will
allow the researchers to simulate aspects of a honeybee’s brain by using a
large group of paralleled desktop PCs. In other words, put together
enough desktops and you can approximate some of the functions of a cluster of
supercomputers, but at a fraction of the cost.
No matter
how much or little money and equipment are involved, this part of the Robo-Bee
project, building the bee’s mind, is an even more formidable challenge than
building a robotic insect that just flies. The mind of even an insect is
breathtakingly complex, but the Green Project researchers are not trying to
tackle the replication of the honeybee’s entire brain. Instead, they are
focusing on only two functions: vision and the sense of smell.
Researchers
are attempting to develop cognitive models of sight and smell. To
duplicate even part of an actual bee’s brain, you need to study an actual bee
or, at least, work with someone who has. That someone is Martin Giurfa of
Toulouse, “an expert in all aspects of bee brain anatomy, physiology, and bee
cognition and behavior.” The ultimate goal is a robotic bee that can
detect particular odors or particular flowers. But, more immediately, the
researcher are hoping to develop computer models of these processes that,
someday, will be downloaded directly into the “brain” of a robotic bee.
However,
the description above understates and ambition of one aspect of this
project. The researchers are attempting to develop models with true
artificial intelligence. That is, they are attempting to develop a
computerized intelligence that will allow a robotic honeybee to act
autonomously. Put yet another way, these robotic bees would have the
cognitive ability to perform certain basic tasks without pre-programmed
instructions. In other words, these
robotic bees would be able to think.
These types
of cognitive models are more than a
few steps beyond simple programming. But why bother? Why do researchers need models of cognitive
processes associated with vision and smell?
Couldn’t they develop an artificial intelligence without these senses?
The surprising answer is . . .
probably not.
“EMBODIMENT” – SOMETHING TO
THINK ABOUT
What do we
think about? We think about what we see, smell, hear, feel and
taste. Could a human intelligence ever be “designed” without senses and
sensory input? No.
So, in
order to develop a real artificial intelligence — an intelligence that thinks
-- that intelligence must be “embodied” with those senses that provide the
necessary sensory input (something to think about).
Simply put,
the concept termed “embodiment,” applied to robotics, “holds that any true
artificial intelligence is impossible unless the robot has sensory and motor
skills that connect it to the world.” In other words, without a sensory interface
with an environment, cognitive intelligence, as we know it, wouldn’t exist.
CONCLUSION
So, if Robo-Bee is going to pollinate, Robo-Bee will have to
be able to think.
POSTSCRIPT
As I made
my way through each layer of technological development necessary to launch that
first Robo-Bee into the meadows and fields, I couldn’t help but notice a
strange contrast.
The goals
of Harvard’s MAV project include perfecting an insect-sized robot and that
flies like an insect. This finished Robo-Bee
would even be able to fly in swarms with other Robo-Bees -- adapting its
responses to changing conditions. But reaching
these goals, alone, wouldn’t produce a practical robotic honeybee – one that
could go into agricultural fields and pollinate.
The Green
Brain Project, on the other hand, picks up where Harvard’s MAV project leaves
off tackling the daunting task of perfecting the artificial intelligence needed
to make the robotic bee do what a bee does.
The
contrast is in the funding.
Harvard’s
MAV project is quite well funded and has allowed researchers to work their way,
slowly, around and through an amazing number of obstacles. And, a lot of slow progress is still ahead.
The Green
Brain Project is certainly funded, but not anything like Harvard’s MAV
project. The Green Brain Project has to
run desktop PC’s in tandem instead of being able to afford the computer
“muscle” ideally required. The accommodations
are adequate, but considering the importance of the cognitive functions to a
working robotic bee, why the “dip” in financial interest when it comes to the
Green Brain Project?
Maybe the
answer can be found in the story of the mysterious Robo-Fly.
Thursday 18 September 2014
GCLM5444HOxenia
The Next Post Is Coming On: 4 October 2014
