RealTechNews.com - Independent Tech

Friday, October 29, 2004

Brain Cells in a Dish Fly Fighter Plane
Friday's Guest Rant from Cyrus, Our Intern: Yesterday it was genetically modified cats, and today rats are flying planes. Researchers at the University of Florida extracted neurons from rat embryos then placed them on a grid of gold electrodes. The cells grew, creating a "live computation device". The completed grid of over 25000 neurons was then attached to a flight simulator, where algorithms written by the researchers were used by the cells to determine the status of the plane in different conditions, issuing corrections if the brain "felt" they were needed.

I like this in one way, as anything that can save a human life or save us time or money is a great advancement in technology. However, I am repelled by the fact that we are close to entrusting things to "rat brains" that were previously handled by humans. It seems that we are just making our own species obsolete, making it so that we are never needed, never required. Also, how effective can any algorithm written by a human truly be? I would hate to be the person with my life on the line, when the "rat brain" makes the wrong decision. Our brain is not perfect, but I would rather lose my life to a real human mistake, rather than spending my time in the afterlife wondering if my life would have been saved if the rat had been running Linux instead of Windows Rat.

While advancements such as this are welcome, we need to take the time to really think out the consequences and decide if we truly want to trust something important to a rat´s brain controlled by algorithms written by fault-prone humans.

What do you think?

# Permalink Posted at 2:37 PM 4 comments Email this Link

Comments on this Item:

To be honest, I am not quite certain as to what your issue is with exactly. The fact that rat brains were used or that algorithms written by people were used. The fact is, most modern aircraft to one extent or another do fly themselves based upon control algorithms, the only difference being that these programs are executed by silicon circuits. Modern commercial aircraft such as the Boeing 767 are quite capable of taking off, flying to a destination and landing without a human pilot ever touching the controls. In point of fact most commercial flights are flown on autopilot, with the human pilot only intervening durring landings and takeoffs. As for high performance military aircraft, without the use of computers making thousands of minute corrections every minute, they would be impossible to fly as their designs are inherently unstable. Examples being the F117A, B2, and the F22. Most of the algorithms used are what I think are called fuzzy logig sysmtems, where a digital system tries to replicate an analog response, or basicaly trying to replicate a natural neuronet type system. Now it seems to me the rat neurons, or neurons of any creature would be better at fuzzy type logic than any solid state device. After all, we've only been working with electronics for some odd fifty years or so whlile nature has been working on neurons for several million.

# posted by Anonymous 11:09 PM

You draw a great point with pointing out that many planes are autopiloted, I just draw my issue with the fact that when we get into "fuzzy logic" we are setting ourselves up for a problem. When I am in a Boeing 777 I trust that autopilot, because it knows the right answer. When we deal with fuzzy logic, such as the rat brains or to an extent ours, I fear that we are setting ourselves up for trouble. First, its making more free decisions, not following set programing. That could mean the "fuzzy logic" is brilliant or it could screw up and kill us. Military Aircraft is corrected thousands of times a second, but not many people know that it is rare for an aircraft carrier to go out on patrol and come back with all of its pilots alive. These fuzzy logic systems also cost lives in Iraq, as they are used in the autopilot systems of our attack helicopters and we are all aware of how often those crash. While nueron systems could improve this, could it not also make it worse?

I just want to say that for me, this is a hot subject, just like cloning cats or Intel or AMD. If we don´t discuss it, who knows where we could end up, and even if it is a good advance (which I think it is) it could easily slip out of our fingers. Hate to sound SF novelish, but I went to I´Robot and that is the sort of thing we want to avoid.

# posted by Cyrus 9:12 AM

I think you missed the point Cyrus.

I can't imagine that the experiment was intended to somehow introduce rat brains into the IT marketplace. I can imagine that it was intended to learn how brains work. I guess it was a poor choice of programs to use, but maybe they didn't trust any of the current benchmark suites to give consistent results.

Pretty cool concept when you think about though: replacing faulty organic circuits with silicon - Scary to think about BIOS revisions though.

PS: for the record - planes like the F-16, F-117, and F-22 really don't care which direction they fly in. They fly in the direction they do because (most of the time) the engine is pushing them that way.
The B-2 is using pretty much the same platform/control system that was used on the YB-47/49 (1930's Tech), but was updated to relieve the two man crew (I can't imagine doing a 24hr duty day in that thing). This is also why nearly all aircraft that fly in the flight levels use autopilots, it's way too much work to "hand fly" an airplane above FL300 for more than a few minutes.
Ultimately, it will be a very long time before ANY vehicle carrying humans will have a fully autonomous control system (ya gotta have someone to point the finger at).

WRM

# posted by Anonymous 11:19 AM

Like the others, I don't see that the "rat brain" part of it makes it any more dangerous than a silicon based computer making fuzzy logic decisions in flying a plane. Once again, case in point is commercial aircraft. I saw a BBC show where they had a camera and interviewer inside the cockpit of a British Airways 747 on a flight from London to New York. The autopilot handled everything from takeoff to landing. The only thing the human pilot did was turn on the reverse thrusters after the plane had touched down. So this kind of stuff is happening already with silicon. Are rat brains any less effective at this than silicon? No idea. But the algorithms that the "fallible humans" are writing are not going to be that much different, given that they are working on the same task (keeping a plane in the air will take the same steps regardless of what material you make the computer out of - the low level code may be different, but sea level is still sea level, etc.)

# posted by Anonymous 8:34 AM

Post a Comment