Rat brain flies jet

Capt'n Midnight

Military money being spent on AI again..
http://www.theregister.co.uk/2004/12/07/rat_brain_flies_jet/

Florida scientists have grown a brain in a petri dish and taught it to fly a fighter plane.

Scientists at the university of Florida taught the 'brain', which was grown from 25,000 neural cells extracted from a rat embryo, to pilot an F-22 jet simulator. It was taught to control the flight path, even in mock hurricane-strength winds.

"When we first hooked them up, the plane 'crashed' all the time," Dr Thomas DeMarse, an assistant professor of biomedical engineering at the University of Florida, said. "But over time, the neural network slowly adapts as the brain learns to control the pitch and roll of the aircraft. After a while, it produces a nice straight and level trajectory."

The brain-in-a-dish was DeMarse' idea. To produce it, 25,000 rat neurones were suspended in a specialised liquid to keep them alive and then laid across a grid of 60 electrodes in a small glass dish.

The cells at first looked like grains of sand under the microscope, but soon began to connect to form what scientists call a "live computation device" (a brain). Electrodes monitor and stimulate neural activity in this network, allowing researchers to study how the brain processes and transfers information.

The scientists hope that their research will lead to hybrid computers with organic components, allowing more flexible and varied means of solving problems.

One potential application is to install living computers in unmanned aircraft for missions too dangerous for humans. It is also hoped that further advances will help in the search for cures for conditions such as epilepsy, The Age reports.

"The algorithms that living computers use are also extremely fault-tolerant," Dr DeMarse said. "A few neurons die off every day in humans without any noticeable drop in performance, and yet if the same were to happen in a traditional silicon-based computer the results would be catastrophic."

The US National Science Foundation has awarded the team a $500,000 grant to produce a mathematical model of how the neurons compute.

Note this is only 25,000 cells - modern microprocessors would have about three orders more transistors.

John

Capt'n Midnight wrote:

Military money being spent on AI again..
http://www.theregister.co.uk/2004/12/07/rat_brain_flies_jet/


Note this is only 25,000 cells - modern microprocessors would have about three orders more transistors.

I don't know much about transistors but 25,000 neurons would have on average about 250,000,000 synaptic connections and this is a conservative estimate (10,000 synapses each, most neurons range from the 1,000 to 10,000 mark but many types of neurons can have up to 30,000 synapses). That's a lot of connections.

And saying "only 25,000" isn't putting much faith in neurons, the majority of life on earth (including thousands upon thousands of species that use flight) have far less than this. And consider that neurons can learn (nervous tissue is highly plastic) I'd say if they can get this working it would be an extremely impressive feat (albeit slightly worrying in my opinion).

Capt'n Midnight

http://www.nervenet.org/papers/NUMBER_REV_1988.html

The total number of neurons in the central nervous system ranges from under 300 for small free-living metazoans such as rotifers and nematodes (e.g., Martini 1912), 30–100 million for the common octopus and small mammals such as shrews and mice (Young 1971, Campbell & Ryzen 1953, Williams 2000), to well over 200 billion for whales and elephants. Estimates for the human brain range between 10 billion and 1 trillion.

An organism like c. elegans http://elegans.swmed.edu/ might only have 300 neurons - but then again it has less than 1000 cells. I can't find any reference to any insect with less than millions of neurons. Out of curiousity is the interconnection level average accross all species or just humans ?

Interconnections in cpu's can be very high - most gates being linked to the clock and probably some of the data/address buses and in memory circuits there are lots of connetions too. I'd hate to have to draw out a 64 bit carry look ahead adder.. But as you point out they are static being hardwired.

In brains / ganglions most of the complexity comes from the ability of neurons to form and unform connections over time, the triggering of a neuron depending on a weighted average of the others it is connected to that have triggered getting past a threashold. About a decade ago I saw work on replication that in silicon but haven't heard much since.

If you remember the cat they were remote controlling by generating hunger signals in it's stomach this is worrying.. They only used 60 electrodes, generating many times more is very easy think of an LCD panel with millions of "eectrodes" combined with CCD technology you have a neat I/O system that could easily cover the surface of the neurons. And 25,000 neurons use less power and are lighter than the corresponding electronics, reckon we'll see lots of proposed civilian uses for this technology that somehow fail to materialise while the real work goes on behind closed doors.

vibe666

it's impressive at any rate. could well open up lots of doors in the future. i wonder if the cource of the neurons would have an impact on their effectiveness.

surely one of those scientists could spare a few thousand to give it a go.

elivsvonchiaing

Think when Moore's law finally collapses, this is where we need to go. Though I can forsee robots with depression

Capt'n Midnight

elivsvonchiaing wrote:

Think when Moore's law finally collapses, this is where we need to go. Though I can forsee robots with depression

Marvin the paranoid android :
"life, don't talk to me about life, lothe it or ignore it, you can't like it. "

elivsvonchiaing

Seems like a pointless post... (no rep system anymore) But yes indeed
'twas from hitchiker's guide to the galaxy! Brill book/series! Imho!

vibe666

elivsvonchiaing wrote:

Seems like a pointless post... (no rep system anymore) But yes indeed
'twas from hitchiker's guide to the galaxy! Brill book/series! Imho!

funny, i was going to post a marvin quote too, but decided not to.

elivsvonchiaing, you're just not geeky enough to be hanging around in these parts.

don't worry Capt'n, I'd have given you a positive to counter the neg anyway.

where did that go by the way? I was just getting enough spread around to berate richard dower again (he does seem to need it after all), when all of a sudden it was gone. :mad:

John

Capt'n Midnight wrote:

An organism like c. elegans http://elegans.swmed.edu/ might only have 300 neurons - but then again it has less than 1000 cells. I can't find any reference to any insect with less than millions of neurons. Out of curiousity is the interconnection level average accross all species or just humans ?

Interconnection would be very similar in rats and humans, the cell structures are basically the same. Of course the overall structures of the brains are different but not that different. We have everything they have just our brains have been pimped

And as regards insects, as far as I remember a lot of ants have only about 10,000 neurons (I think the queen has much higher) and the male ants have flight during the mating season.

elivsvonchiaing

vibe666 wrote:

elivsvonchiaing, you're just not geeky enough to be hanging around in these parts.

Agree with the rep thing. Only reason for this pointless post! Beg to differ I can be geeky when I put my mind to it Of course there are others who may pour cold DHMO on this

vibe666

elivsvonchiaing wrote:

Of course there are others who may pour cold DHMO on this

careful now, I've heard nasty things about that DHMO stuff. It's a lethal; chemical that's lilled millions of people from what I understand.

just be safe and don't go pouring it on anything.

John

The scary thing about DHMO is that it's found in 100% of all water supplies

vibe666

:eek:

Victor

Screw that I can drive a car in my sleep.*

* Of course I can't drive while awake :eek: