Head Cheese Gone Wild
I was plenty pleased when little porridges of cultured neurons took their first baby steps towards running flight simulators or operating robots in the lab; I was downright smug when folks noticed that I’d got there first. Now, though, researchers from the Missouri University of Science and Technology are planning on putting head cheeses in charge of real-world power grids in half a dozen countries, including China and Mexico (but not including, interestingly enough, the United States). According to this article, “…these networks could control not only power systems, but also other complex systems, such as traffic-control systems or global financial networks.”
Traffic control systems. Financial networks. Being run by meaty neuron networks whose thought processes are, by definition, opaque. For real.
I wrote a trilogy about just this scenario. It did not end well (just ask Kirkus). Maybe someone could pass a copy on to this Venayagamoorthy dude.
Next up, two papers in today’s issue of Science: one on the evolution of religious belief, the other on the perception of imaginary patterns under conditions of perceived helplessness. These dovetail nicely with some slightly staler findings on the arrogance of stupid people, the inherent fear responses of political conservatives, and last night’s competing North-American neocon/centrist debates. But I have to actually watch those debates before I blog on that. (I was out at Don Giovanni last night. I didn’t even know that they had dry-ice smoke machines in 1787…)
Being run by meaty neuron networks whose thought processes are, by definition, opaque. For real.
Just like our governments.
Being run by meaty neuron networks whose thought processes are, by definition, opaque. For real.
Or the American “banking” industry.
Hmm… head cheese *experiments* in developing countries, then presumably when the grey goop is ‘smarter’ it becomes good enough for use in the *US* ??!(shudder).
Also re the linked article: How’d they go about punishing/rewarding the cells? The article just says ‘the neural network slowly learns..’ – Learns what? That it get zapped with a squirt of bleach if the plane crashes? I wonder…
The upshot of this tech is that it might pave the way for an interesting lateral development.
I’m speaking of course about direct brain/device communication which is a huge slab of concrete on the path to singularity.
If they really want to crank it up they just need to figure out how to put a cat brain into a killer robot and wire it so that it sees targets as…well, targets.
There’s potential there.
Leona said…
The article just says ‘the neural network slowly learns..’ – Learns what? That it get zapped with a squirt of bleach if the plane crashes? I wonder…
I’ve been wondering that myself for years. I’m given to understand that when conventional nets in silico get something right, the associated pathways get their resistance amped down, thus lowering the threshold for their activation in subsequent runs — but how you'd do that to a real neuron pancake I don't know. The bleach wouldn't even work, since brains got no pain sensors.
Anyone?