Rat brained robots and reportersFriday, 15 August, 2008
As news stories have been reporting for the last 36 hours or so, researchers at the University of Reading in the UK (home of the Project Cyborg, led by Prof. Kevin Warwick) have made a robot controlled by rat neurons.
But many reporters, being fundamentally rat-brained themselves, have made the mistake of saying silly things like these:
To build the cyborg, Warwick, his colleague Ben Whalley and their team dissolved a fetal rat’s brain. They then put the free-floating brain cells, or neurons, into an electrode-ringed Petri dish, where the cells quickly reassociated with each other and began randomly firing electrical signals. Electrical pulses sent through the ring of electrodes calmed down the neurons, and after a while the cells began to “learn” patterns among the external pulses. [from FOXNews]
Now the reason there are quote marks around the word “learn” is because it didn’t really happen.
It is true that a network of 300,000 murine neurons (i.e. rat brain cells) are controlling that robot through 80 electrodes. It’s true that the cells, which were dissociated in solution, spontaneously formed that network (which was, in my opinion, the most exciting part). But it is not a robot controlled by a rat’s brain, for these are only brain cells – admittedly forming a primitive brain, but not a brain anything like that of a rat.
And it is most certainly not true that these neurons learnt how to control the robot – the researchers set up the robot and electrodes so that whenever the sensors detected a wall, it sent a 1 volt pulse to the neurons in the dish. And the neurons, being fully functional, were elicited into giving off an action potential. Just like neurons do when hit with an electrical stimulus of 1 volt (which is massive for such small cells, by the way). And this action potential, after it had been transmitted through a few neurons, was recorded and used (by a computer) to make the robot turn away from the wall. It is not the neurons in the dish that learnt to control the robot, but the computer that is reading the activity of those neurons (and the programming of that computer by human brains).
From a neuroscience perspective, the neurons have not learnt much. They have probably strengthened their connections to the electrodes, but it is a far cry from learning to control the robot. Those 300,000 neurons aren’t really doing much that the 302 neurons of the nematode worm C. elegans couldn’t do – in fact, they are probably doing far, far less.
Aside from the fact that the neurons formed a network spontaneously, I am unimpressed.
UPDATE: Steve Potter, a neuroscientist from Georgia Institute of Technology has commented on the blog Neurophilosophy, saying:
I see nothing new here beyond what we and others have been doing for the past 5 years. Believe what you read in peer reviewed papers.
So that explains why I, being a neuroscience student, wasn’t at all impressed.