Aug 14 2008

Robot with a Biological Brain

Published by under Uncategorized
Comments: 32

I enjoy science fiction partly because it can be a thought experiment on the potential course of future technology. A common sci-fi theme is the merging of man and machine and the blurring of distinction between the two. Clearly this is a process that has already begun, but even the most thoughtful futurists cannot tell where this will lead with anything but the broadest brush strokes.

In the campy sci-fi flick Saturn 3 (1980), the character played by Harvey Keitel builds a robot (Hector) with a modified human brain as it’s CPU. Keitel trains the robot partly by imprinting his own brain patterns onto it. As he is a dangerous criminal psychotic, antics ensue.

The prequels to the classic Dune series, written by Frank Herbert’s son, Brian, along with Kevin J Anderson, the primary enemies of humanity are cymeks – immortal human brains that can inhabit their choice of robotic bodies – from battle armor to space ships.

In 2061: Odyssey Three, Arthur C Clarke paints a future where every human is equipped in infancy with a “brain cap.” This is a super-computer that fits nicely over the skull, sending electrodes down into the brain in order to seamlessly interface with it and greatly expand human intellectual capacity.

In Stephen R. Donaldson: The Gap Series (1990-94), a major plot element is the “zone implant.” This is a computer device implanted into the human brain in order to enhance and control its function. They are outlawed because of they offer complete control to whoever holds the remote control to someone’s implant. But they are also portrayed as useful for the treatment of psychological and neurological disorders.

And, of course, in the new Battlestar Galactica series (I highly recommend this series to any sci-fi fan, and even if you are not a sci-fi fan this is simply superb drama) the Cylons have created humanoid robots that are indistinguishable from normal humans but have Cylon AI.

This is just a sampling running the spectrum from brains controlling robots to AI controlling human bodies with brain-AI interfaces in between. At present we are just taking the first baby steps toward whatever version of these fictional futures await us, if any. Time will tell.

Frankenrobot

Kevin Warwick and other researchers at the University of Reading have developed a robot that is entirely controlled by rat neurons. The robot, called Gordon, is the first of its kind. Warkick isolated neurons from rat fetuses then placed them on a 5 inch by 5 inch array of 60 electrodes. He reports that the neurons with a day started to synapse together, forming a network. The array outputs to the robot’s motors and receives stimulation from it’s sensors. Within a week the neurons start to exhibit brain-like activity and then Warwick can begin “teaching” it by having the robot explore it surroundings. Most of the learning is automatic, based upon feedback from the receptors. But Warwick also reinforces the learning by using drugs to reinforce or inhibit certain neural pathways.

What Warwick and others are doing is building  brain from the ground up. This is a great way to see how neurons behave and how complex behavior emerges from the basic properties of neurons. They seem to spontaneously network together and form useful pathways based upon the neural feedback they get. This is likely to be a very fruitful avenue of basic neuroscience research.

This research may also lead to the development of biological computers. Extrapolating from this current research, it is not hard to imagine growing neurons on a complex matrix that results in a powerful parallel processor – a brain, of sorts, but designed to perform computing functions and interfacing with electrodes attached to a familiar computer interface. This technology may not be practical for desktop applications (and this is where predicting future technology becomes difficult to impossible) but there may be applications for which massive parallel processing, even with relatively large and slow neurons, has advantages over silicon.

One potential advantage of neurons over silicon is the former’s ability to learn and adapt. Gordon, in part, trains itself just by moving around and bumping into things. Neural pathways are adaptive, not etched in silicon.

These experiments also have the potential to increase our ability to interface neurons and computers – a necessary technology if we are going to have our “brain caps” one day.

Another aspect of this research worth discussing are the ethical considerations. Warwick largely avoids controversy by using rat neurons (rather than human neurons), but you can see where this technology is going. I wonder how long it can fly under the radar before the protests begin. I foresee huge ethical and social conflicts in the future as this technology develops. Is Gordon “alive”? Is his matrix of neurons a “brain”? If the same were done with human neurons would that make Gordon a human? How large and complex will such a network of neurons have to be before it is considered a being? or to have the rights of a human? Going the other way, how does the introduction of computer AI into a human brain affect a person’s status as a person?

The waters are muddied and the lines are blurred, and if this technology develops they will increasingly be so until there is such a smooth continuum from human to machine that no objective lines can be drawn.

Many sci-fi works have explored these themes as well.  We can imagine nightmare scenarios like the Matrix in which humans are enslaved in a virtual world by the very machines we created. But we can also envision a bright future, like Clarke’s 2061, in which machine interfaces give humanity more power over themselves and their environment.

The reality is likely to be messy and complex, neither our worst nightmares nor our brightest dreams. One thing is for sure – it is likely to be nothing like we can imagine today.

32 responses so far