A collaboration art project between the Potter Lab at Georgia Tech and SymbioticA Research Group recently grabbed the attention of the likes of Nature and Wired. The project is called MEART (official site). In short it is a robot arm that draws pencil drawings on paper, controlled by real living rat neurons. And what is more, the “brain” is in Atlanta, USA, while the “body” is in Perth, Australia, communicating over the Internet!
Images from a webcam near the robotic arm, taken every morning, are used as stimulation to the neurons that are grown on an electronic chip. In short, the image is changed into a 60 pixel thumbnail that correlate to 60 electrodes that stimulate the neurons. This image is compared to what the robotic arm has already drawn on the paper, giving highest value to spaces that are dark in the photo and blank on the paper. The neurons process this information and after a short pause the electric signals on the same electrodes is measured again. This measurement is then used to guide the robotic arm. This way the pictures taken by the web cam affect the robot arm’s drawing.
This project interests me for a variety of reasons. The neuron-chip interface is quite striking. The people at Potter Lab have used that one before in another ground-braking project where they built a rat-brained robot that found its way in mazes. Potter’s people are no doubt pioneering in this field.
Secondly the distance between the “brain” and the “body” is interesting. It is of course typical Internet traffic and nothing special about it as such, but the wetware aspect of it almost automatically leads to philosophical ponders about the mind-body problem and whether something residing in two places can still be considered one “organism” (which of course relies on the question whether it is acceptable to call the MEART robot an organism).
One thing bothers me however. In the case of the maze robot, it had a clear goal and a feedback loop telling the “brain” whether it was progressing or not in finding its way through the maze. In the case of the MEART robot, there is no such feedback. The audience does not tell the robot whether they like the images or not, the robot is not working towards anything in particular. Not that it is in any way necessary to create the art, but from a scientific standpoint it does not drive the “brain” towards anything. If you’d grow another such brain, it would probably behave in a totally different way and any changes in the neural connections during the process are totally arbitrary. Odds are that you could throw the values from the webcam images into any complex system and get similar results, all that the neurons do is creating unexpected values from the input variables. Analysing how the neurons behaved will no doubt be of scientific value but it would have been more interesting to see if the MEART robot could have gotten more “creative” or “intelligent” had there been such feedback.
Maybe we will see such experiments soon.
The MEART paper (quick overview, pdf-format)