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 

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

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)