spacer

Do Robots Dream of Electric Sheep?

In his famous novel "Do Androids Dream of Electric Sheep?", author Philip K. Dick delves into the dark realms of what, if anything, separates humans from the artificially constructed androids that share their world. In the book, police have to resort to using more and more subtle tests to distinguish between humans and the rapidly advancing androids. Eventually one is forced to wonder whether the hardware matters all that much.

Most people would say that there is a vast difference between living things and machines. A toaster that autonomously toasts our bread is still a world away from even the lowliest of bugs. The distinction seems as obvious as black versus white. But, if you begin to explore the regions between these extremes the distinctions become less clear. I see already that there are significant areas of overlap between the most complex machines and the simplest organisms. The smartest machines already posses abilities that exceed the simplest lifeforms.

For some people these arguments are moot. Lifeforms can reproduce and machines/robots cannot. A compelling argument until you consider viruses. These molecular assemblies straddle the line between machines and organisms. Individually they are unable to reproduce like a machine. Yet, they are able to commandeer the reproductive machinery of cells to produce copies of themselves and thus reproduce and evolve. Machines also share this mechanism of reproduction. If I publish a enticing design for a robot and another hobbyist builds a copy, hasn't the robot reproduced by commandeering the hobbyist? And so the line blurs and begins to fade away.

The word robot encompasses a vast class of devices ranging from preprogrammed mechanical arms in factories, to the R/C vehicles in 'Robot Wars', to fully autonomous 'self-driving' cars that Google builds. For me, robots should be autonomous. The factory arm or R/C robot have no possibility to do anything unexpected. Although impressive, they are essentially extensions of the human who programmed their movement or wields the R/C transmitter. If a machine has no chance to do something unexpected, it fails my robot test.

My 2011 March Break Robot falls below my threshold for a real autonomous robot. While it could 'line follow' or 'sumo wrestle' it had no understanding of what it was doing. It's behaviour was a side effect of the robot following a set of predetermined actions; actions that I the programmer had programmed into it. As such, the robot was no more autonomous that an R/C Robot. It's a subtle but important difference for me.

Living organisms have motivations. These are what drive behaviour. When the organism is hungry it searches its repertoire of behavoirs and finds that eating something satisfies this motivator. Now the organism is motivated to find something to eat. What behaviours does it know that will find something to eat? Perhaps hunting, or perhaps returning to a location where it found food before. Perhaps it will start crying because last time that produced food.

To do this effectively some form of memory is required or else past successes are lost and the organism is forever reinventing the wheel. Or, the behaviour is hardwired and the organism becomes more of an organic machine than a lifeform.

My goal, within the limits of the hardware and my abilities, is to give Fitzy and Carraldo a set of motivations and a range of behaviours to choose from and see where it leads. Of course these simple robots have limited abilities, they are virtually blind, and their brains are more simple than a gnat's yet still I remain optimistic.

One motivation I think the robots need is a sense of home. When they wake there will be a beacon nearby that will signal home. Being near home will be comforting.

Another motivation I will give the robots is a sense of boundary. Within their abilities they will have an urge to discover boundaries. To determine extents and to map their environment. This will be useful as it will help them find things that will satisfy other needs.

Balancing this there will be anxiety when their path is blocked. There will be a strong motivation to be able to move freely.

As their batteries empty, they will become hungry and that will motivate them to discover ways of feeding.

They will have a sense of boredom. Surprisingly boredom is a very strong motivator in humans. Performing repetitive series of actions will build boredom. Novelty will become something to seek out. Perhaps a boundary that has changed location, or a new object found.

And finally I want to give them one additional motivation. Scattered throughout their environment will be blocks. They will have an urge to collect these and arrange them near their home. This will be balanced by their need for free movement.

Hopefully these simple motivations will lead to lifely behaviour. Lifely because lifelike implies imitation. If a robot exists in its environment complete with wants and needs isn't it at least as alive as a gnat?

Next Section: The Fitzy and Carraldo Project