Note: Where available, the PDF/Word icon below is provided to view the complete and fully formatted document
Disclaimer: The Parliamentary Library does not warrant or accept liability for the accuracy or usefulness of the transcripts. These are copied directly from the broadcaster's website.
Brainy Bees -

View in ParlViewView other Segments

Brainy bees (30/04/2009)

TRANSCRIPT

NARRATION:

Every day millions of people travel through our cities, airports, bus depots and train stations...

NARRATION:

And the sad thing these days is they have become targets for terrorists. A way to better protect
ourselves would be to have computers connected to security cameras, constantly scanning the crowds
for suspected terrorists. But that's not as simple as you might think.

Dr GRAHAM PHILLIPS:

To spot a suspect in a crowd a computer needs to scan lots and lots of faces. But getting a
computer to see and recognise an individual face is actually very difficult. It just may be we can
learn more about the science of seeing by studying bees.

NARRATION:

Dr Adrian Dyer is a visual researcher who studies how animals process information and learn from
what they see.

Dr Adrian Dyer:

The systems which have been developed, they have a lot of problems if the face is viewed from a
different angle so if someone's walking through an airport and a ╛ view of their face is captured
then systems produce many many false positives and are essentially not very useful for the task
they were designed and there's a lot of hope we might be able to get some solutions from biological
systems

NARRATION:

But why study a bee?

Dr Adrian Dyer:

Insects have a very very constrained brain because they have to fly so it has to be very small and
light. That tends to mean they have very efficient solutions for solving particular tasks.

NARRATION:

A bee has this task solving) ability despite its brain being ten thousand times smaller than a
human brain. And perhaps it's even more remarkable when we look at the world the way a bee sees it.

Dr GRAHAM PHILLIPS:

Now bees don't see faces like we do. They have compound eyes, so each eye is made up of thousands
of much smaller but very simple eyes. Well the scientists have simulated that by putting together
thousands of straws giving you a bee's eye view of the world.

NARRATION:

The device mimics a bee's mosaic-like view. It also shows that objects - such as faces - are much
less sharp when they're further away from the bee.

Dr Adrian Dyer:

It enables us to get a pretty good representation of bee vision cause we were able to match the
resolution of the system to the previously determined behaviour data for honey bees.

NARRATION:

Humans, and primates in general, are very good at recognising faces. In fact, we seem to have parts
of our brain specially devoted to the task. But bees don't.

Dr Adrian Dyer:

So a bee we might see in the garden, she can not recognise a face. The exciting thing is that if
she's taken and given the correct conditioning her brain will learn how to recognise a face.

NARRATION:

To study bees Adrian needs a hive. These foraging bees fly to a feeder about 15 meters away from
the hive.

Dr GRAHAM PHILLIPS:

So you're actually just bringing in wild bees here?

Dr Adrian Dyer:

Yeah, so these bees just elect to come and feed at this site, and they come and collect 5% sucrose
solution. We'll take one of these bees and take it up to a test site and train her how to do a
particular experiment.

Dr GRAHAM PHILLIPS:

So the experiment's up here

NARRATION:

The selected bee is coaxed to the experiment site using a 25% sugar solution.

Dr Adrian Dyer:

So what this experiment is doing here is we're training an individual bee to learn to recognise the
difference between a given target face which we see here and a distractor face which is
perceptually very very very similar. So the bee is taken a short distance away and then we can
randomise the positions and see if the bee can learn to go back... and she's aborted the distractor
there and gone back to the correct target face.

Dr GRAHAM PHILLIPS:

She's not doing badly. They're very similar faces.

Dr GRAHAM PHILLIPS:

Yeah, she's not doing too bad. This bee is still learning so she's made it to high school level but
it's going to take her the whole day to learn the task.

NARRATION:

When the bee has collected enough sugar water she returns to the hive, unloads it and then returns
to the experiment site ready to collect more from the correct target face.

Dr Adrian Dyer:

The stimuli we used for the experiments were taken from standard psychology type tests and we did
that because we knew the faces were of a particularly difficult nature even for some humans. And so
they were able to learn very very similar types of faces so they would easily tell your handsome
face from my face

NARRATION:

But incredibly, recent studies show that bees can not only recognise an individual face, they can
recognise a rotated view of it which they have never seen before.

Dr Adrian Dyer:

There's been a study published in 2007 2008 showing that new born human babies could recognise
rotated faces so we wanted to know if a very general system could do it.

NARRATION:

Different groups of bees were trained to recognise different views of faces.One group was trained
to recognise a face front on, while another to recognise a face rotated to 60 degrees.When these
bees were shown the face rotated to 30 degrees neither group could recognise it.But the main test
group was trained to recognise both a zero degree view and a 60 degree view. When tested, they were
able to recognise a 30 degree view of a face.

Dr Adrian Dyer:

They learn to solve this very difficult face recognition task through a method of image
interpolation or averaging different views that they've seen

NARRATION:

The US air force is so interested in how bees learn to recognise faces they've helped fund some of
Adrian's research.

Dr Adrian Dyer:

No one ever knows for sure what will happen out of a set of experiments but bees in the past have
provided some interesting solutions for computing. The fact that we have some evidence that the bee
brain can recognise faces means the bees might have some clever tricks or novel solutions which
might be of benefit for teaching computing systems how to be very clever at recognising faces.