Science News

Embodiment, Computation And the Nature of Artificial Intelligence

The notion of intelligence makes no sense without a broader view of computation, argues one of the world's leading AI researchers

One of the buzzwords in artificial intelligence research these days is 'embodiment', the idea that intelligence requires a body.

But in the last few years, a growing body researchers have begun to explore the possibility that this definition is too limited. Led by Rolf Pfeifer at the Artificial Intelligence Laboratory at the University of Zurich, Switzerland, these guys say that the notion of intelligence makes no sense outside of the environment in which it operates.

For them, the notion of embodiment must, of course, capture how the brain is embedded in a body but also how this body is embedded in the broader environment.

Today, Pfeifer and Matej Hoffmann, also at the University of Zurich, set out this thinking in a kind of manifesto for a new approach to AI. And their conclusion has far reaching consequences. They say it's not just artificial intelligence that we need to redefine, but the nature of computing itself.

The paper takes the form of a number of case studies examining the nature of embodiment in various physical systems. For example, Pfeifer and Hoffmann look at the distribution of light-sensing cells within fly eyes.

Biologists have known for 20 years that these are not distributed evenly in the eye but are more densely packed towards the front of the eye than to the sides. What's interesting is that this distribution compensates for the phenomenon of motion parallax.

When a fly is in constant forward motion, objects to the side move across its field of vision faster than those to the front. "This implies that under the condition of straight flight, the same motion detection circuitry can be employed for motion detection for the entire eye," point out Pfeifer and Hoffmann.

That's a significant advantage for the fly. With any other distribution of light sensitive cells, it would require much more complex motion detecting circuitry.

Instead, the particular distribution of cells simplifies the problem. In a sense, the morphology of the eye itself performs a computation. A few years a go, a team of AI researchers built a robot called Eyebot that exploited exactly this effect.

What's important, however, is that the computation is the result of three factors: simple motion detection circuitry in the brain, the morphology or distribution of cells in the body and the nature of flight in a 3-dimensional universe.

Without any of these, the computation wouldn't work and, indeed, wouldn't make sense.

We've looked at examples of morphological computation on this blog in the past (here and here for example). And Pfeifer has been shouting from the roof tops for several years, with some success, about the role that shape and form play in biological computation.

But today he and Hoffman go even further. They say that various low level cognitive functions such as locomotion are clearly simple forms of computation involving the brain-body-environment triumvirate.

That's why our definition of computation needs to be extended to include the influence of environment, they say.

For many simple actions, such as walking, these computations proceed more or less independently. These are 'natural' actions in the sense that they exploit the natural dynamics of the system.

But they also say it provides a platform on which more complex cognitive tasks can take place relatively easily. They think that systems emerge in the brain that can predict the outcome of these natural computations. That's obviously useful for forward planning.

Pfeifer and Hoffmann's idea is that more complex cognitive abilities emerge when these forward-planning mechanisms become decoupled from the system they are predicting.

That's an interesting prediction that should lend itself to testing in the next few years.

But first, researchers will have to broaden the way they think not only about AI but also about the nature of computing itself.

Clearly an interesting and rapidly evolving field.

Ref: arxiv.org/abs/1202.0440 :The Implications of Embodiment for Behavior and Cognition: Animal and Robotic Case Studies

WRITE

Creepies 'n' Crawlies

The best of the rest from the the Physics arXiv this week

Spatiotemporal Features Of Human Mobility

Prime Numbers, Quantum Field Theory And The Goldbach Conjecture

Gerbert Of Aurillac: Astronomy And Geometry In Tenth Century Europe

A Multiple Of 12 For Avogadro

The Direction of Gravity

WRITE

The Secret Of Ant Transportation Networks

Just how ants create the highly efficient network of trails around their nests has never been fully understood. Now researchers think they've cracked it

Among the most impressive transportation networks on the planet are the complex trails that ants create around their nests. These networks arise through the ants' exploration of their environment and end up channelling the distribution of food for the colony and the daily movements hundreds of thousands of individuals.

What's more, these networks aren't just a random criss-crossing of space. Instead, they are a highly efficient solutions to the problem of searching and transporting food. Various groups have created ant-like foraging algorithms to do other types of virtual exploration.

One question that has fascinated biologists is how ants build these networks. They've known for some time that ants leave small deposits of pheromones as they travel and that other ants follow these trails, leaving their own deposits. This increases the concentration of the pheromone, strengthening the trail.

But the precise algorithm that governs the way ants respond to pheromones has been harder to pin down. Many experiments show that a trail can only be reinforced if ants have a disproportionately higher probability to follow a trail with higher pheromone concentration.

Biologists have always assumed that this disproportionate response means ants must have a non-linear response to the chemical. In other words, an ant's tendency to turn towards a pheromone deposit is related in a non-linear fashion to the concentration.

But that seems to conflict with one of the great triumphs of experimental biology--Weber's Law, which relates the perceived intensity of a stimulus to its physical magnitude. Biologists know this holds for the human perception of many stimuli, such as the intensity of sound, and have also verified it in many insects. So why not in ants?

Today, Andrea Perna at the Complex Systems Institute of Paris Ile de France and a few pals, resolve the issue. These guys have developed an entirely new way to image pheromone trails which allows them to study ant response to pheromones in more detail than ever before.

They say the structure of ant trails can be entirely explained if the ants's response to a pheromone droplet concentration is linear. "One ant will turn to the left in proportion to the diﬀerence between the pheromone it has on its left side and the pheromone on its right," say Perna and co.

They also point out that this is exactly what Weber's law predicts.

So where does the non-linearity required to create trails come from? Perna and co say that ant behaviour is inevitably noisy. "We show that the required non-linearity does not reside in the perceptual response of the ants, but in the noise associated with their movement," they say.

That's a fascinating result because it reveals how complexity in nature forms with the simplest of inputs.

And it clearly has implications for the study of other complex structures that ants create, such as their nests. Just how ants create these huge vibrant structures has long puzzled biologists.

Perna and friends hint at an answer in their conclusion. "We can imagine that other collective phenomena, such as group decision-making, could also be founded on coupling between Weber’s Law and simple feedback mechanisms."

In the case of nests, this mechanism would have to operate in three dimensions rather than two. But that shouldn't be too much of a challenge. Perhaps a problem that a relatively simple computer model could help solve.

Ref: arxiv.org/abs/1201.5827 :Individual Rules For Trail Pattern Formation In Argentine Ants (Linepithema Humile)

WRITE

Mobile Phone Data Reveals Human Reproductive Strategies

The pattern of calls and texts between humans reveals how women invest more heavily in their main relationship than men; and how this changes as they age

Various studies have shown that the frequency of contact between individuals is a reliable indicator of the emotional link between them. So it should come as no surprise that the data from mobile phone calls is a potential treasure trove of information about the social lives of humans.

But analyses of this data so far have been distinctly unspectacular. For example, the location data associated with phone calls has revealed various new intricacies in the movements of commuters. Interesting but hardly jaw-dropping.

That is set to change with the work of Vasyl Palchykov at the Aalto University School of Science in Finland and a few buddies including a couple of old hands in the form of Albert-László Barabási at Northeastern University and Robin Dunbar at the University of Oxford (of Dunbar's number fame).

These guys have got hold of a corpus of mobile phone data relating to calls between 1.4 million women and 1.8 million men in an unspecified European country. Between them, these phone subscribers made almost 2 billion calls and sent almost half a billion text messages. In addition to the gender of each subscriber, Palchykov and co also managed to get their age as well.

That's significant because it allows them to study not allow the pattern of calls between genders but the way this changes with age.

They began by taking each subscriber and determining the age and gender of the person they werein contact with most frequently, second most frequently and so on. These, they assume, are the 'best' friend, second best friend and so on.

Then, they looked at how the 'best friends' changed as subscribers age. It turns out in general that between the ages of 18 and 40 or so, men and women have best friends of the opposite sex. Palchykov and co assume this reflects the general pattern of mating in society. Second best friends are generally of the same sex at this age.

But they tease the most interesting phenomena out of the fine detail in their dataset. They conclude for example that women are more focused on opposite-sex relationships than men are during the period of their lives when they are reproductively active. That indicates that women invest more heavily in creating and maintaining their relationships than men.

As women age, their attention shifts from their spouse to younger females some 25 years or so younger. That's about equal to a generation gap and Palchykov and co assume these younger females are daughters. This attention shift also seems to equate to the arrival of grandchildren, when the older female again once again begins to invest more heavily.

While older women focus more heavily on younger females, older men maintain an even gender balance in the second best friends, presumably this reflects an equal attention between children of opposite sexes.

What's striking about this is how strongly female relationships are determined by their reproductive cycle. “Women’s gender-biases thus tend to be stronger than men’s, seemingly because their patterns of social contact are strongly driven by the changes in the patterns of reproductive investment across the lifespan,” say Palchykov and co.

Clearly, female reproductive strategies change more explicitly as they age, switching from mate choice to personal reproduction to parental investment and finally grandparental investment, particularly after they reach 40.

However, the most dramatic conclusion from this work is about the pattern of social relationships that play the most important role in society. Palchykov and co say the tendency in the past has been to assume that father-son relationships dominate.

By contrast, “our results tend to support the claim that mother-daughter relationships play a particularly seminal role in structuring human social relationships,” they say.

This difference on the way the sexes invest in relationships is exactly what evolutionary biologists expect. But although previously suspected, it has proved particularly difficult to test. That's why this work is something of a landmark.

Clearly, the ability to study human relationships on such a vast scale opens up a host of new avenues for research in social and reproductive strategies.

In particular, this study looks only at the existence of links between people, not the the directional asymmetries in relationships or who initiates contact. Palchykov and co leave that for another day.

There's a mountain of data ready to be mined on this. And clearly, there's gold in them thar hills.

Ref: arxiv.org/abs/1201.5722: Sex differences in intimate relationships

WRITE

Artificial Braneworlds Made to Collide In Lab

Physicists have simulated two universes colliding inside a metamaterial

One interesting way in which our cosmos may have formed is in a collision between two other universes with extra spatial dimensions called braneworlds.

In this scenario, known as the ekpyrotic model of the universe, our cosmos is just a small four-dimensional corner of a much more complex space.

The ekpyrotic model is interesting because it leads to a flat universe like our own without the need for inflation, the period just after the Big Bang in which our universe supposedly swelled by many orders of magnitude in the blink of an eye.

Without inflation, our universe is just too big to have been formed in a Big Bang-type event. But nobody knows what might cause such a dramatic increase in size. Hence the interest in another way of explaining our existence.

If you're wondering what actually collides in the ekpyrotic version of events, the answer is Minkowski domain walls, essentially the edges of universes with different spatial dimensions.

It's easy to imagine that Minkowski domain walls are entirely theoretical. And indeed they were until now.

Today, Igor Smolyaninov and Yu-Ju Hung at the University of Maryland, in College Park, say they've created Minkowski domain walls in the lab for the first time and even used them to simulate the collision of two braneworlds.

The trick these guys have used is a formal analogy between the mathematics of space time and of electromagnetic spaces. Physicists have known since Einstein's day that it is possible to bend and distort the fabric spacetime—our universe appears to be distorted in just this way on various cosmic scales.

But it is only in the last ten years or so that they've learnt how to do the same on a much smaller scale with electromagnetic spaces. What's triggered this is the development of metamaterials: artificial substances that can bend light in almost any way imaginable.

Smolyaninov is fascinated by one version of this stuff called hyperbolic metamaterial. Inside this substance, monochromatic light propagates in a similar way to massive particles in a Minkowski spacetime, where one spatial coordinate takes on the role of time.

Hyperbolic metamaterials are essentially a series of metal layers separated by a dielectric. Smolyaninov has used this stuff to simulate a number of interesting aspects of cosmology including the Big Bang itself.

The collision between universe's is a variation on this theme. “The “colliding universe" scenario can be realized as a simple extension of our earlier experiments simulating the spacetime geometry in the vicinity of big bang,” he says.

He simulates an expanding universe using concetric rings of gold separated by a dielectric. "When the two concentric ring (“universe”) patterns touch each other (“collide”), a Minkowski domain wall is created, in which the metallic stripes touch each other at a small angle," he says.

Being able to recreate these exotic events in the lab is certainly interesting but it is beginning to lose its novelty. The problem is that this work is not telling us anything we didn't know--the universe behaves the same way inside a metamaterial as it does outside.

What Smolyaninov needs is a way of using his exotic materials to do something interesting. In other words, he needs a killer app. Any ideas?

Ref: arxiv.org/abs/1201.5348: Collision Of “Metamaterial Universes”: Experimental Realization Of Minkowski Domain Wall

WRITE

More Articles...

«StartPrev123Next End»

Page 1 of 3