12 Responses to “Rat neurons learn how to fly a plane”

1. Thought Provoker Says:
   June 4th, 2007 at 8:03 am

   Hi Mike,

   Thanks for the video.

   It looks like this would be consistent with Penrose/Hameroff.

   I have got to run off to work. I would be interested in finding any peer reviewed papers on it.

   Regards,
   TP

Comment by Thought Provoker — June 4, 2007 @ 8:03 am

2. MikeGene Says:
   June 4th, 2007 at 8:40 am

   Hi TP,

   Here is one:

   Demarse T, Cadotte A, Douglas P, He P, Trinh V. 2004. Computation within cultured neural networks. Conf Proc IEEE Eng Med Biol Soc. 7:5340-3.

   In this paper we present three related areas of research we are pursuing to study neural computation in vitro. Rat cortical neurons cultured on 60 channel multielectrode array (MEA) allow the researcher to measure from and stimulate sixty different sites across a small population of neurons grown in vitro. Using this system we can send stimulation patterns into the network and study how these living neural networks compute by measuring its outputs. Our first series of studies uses chaotic control techniques to study the dynamics and potentially control the behavior of cortical network. At the same time, we are beginning to apply a model of computation called the liquid state machine or LSM model developed by Wolfgang Maass to provide a firm mathematical framework from which to proceed with our investigations. Each of these components is integrated into a third area investigating the role of computation and feedback using a real-time sensory-motor feedback robotic flight system.

Comment by MikeGene — June 4, 2007 @ 8:40 am

3. Thought Provoker Says:
   June 4th, 2007 at 9:23 am

   Hi Mike,

   I am having mixed emotions about this.

   Only 60 channels?

   It sounds almost too easy.

   But in my quick Google search there were too many reputable sources quoting the University of Florida and now a research paper.

   If it is a hoax, it's a good one.

   Please understand my hesitation. IMO, the implications of this is very supportive of what I have been conjecturing in the ID and Consciousness thread.

   Too supportive.

   I may be egotistical and arrogant but I'm not so far gone as to believe I am the only one that can put 2 and 2 together and see these connections.

   Is this so threatening to established neuroscientists as to create this large of a blind spot?

   Provoking Internal Thoughts

Comment by Thought Provoker — June 4, 2007 @ 9:23 am

4. Atom Says:
   June 4th, 2007 at 12:35 pm

   Does anyone know how they "punished" the network when the plane crashed, or "rewarded" it when it did not? In other words, how did they coax the network to eventually output the correct signal patterns that corresponded to proper flight?

Comment by Atom — June 4, 2007 @ 12:35 pm

5. Thought Provoker Says:
   June 4th, 2007 at 1:29 pm

   Hi All,

   The key buzz words appear to be "cultured neural networks".

   Here is a paper on with and abstract that starts "A description is given of a functional silicon micromachined device that permits non-invasive, bidirectional, highly specific communication with cultured mammalian neurons."

   Here is an article in the Proceeding of the National Academy of Sciences (PNAS) which includes…"Our experiments demonstrate that cells derived from a population of neural precursors exhibit the characteristics of functional neurons in that the cells form excitatory and inhibitory connections that fire action potentials spontaneously."

   I only had time to look at these two, there are more.

   Provoking Thought

Comment by Thought Provoker — June 4, 2007 @ 1:29 pm

6. Mike Godfrey Says:
   June 4th, 2007 at 1:41 pm

   Hi Mike,
   this is another interesting post.It is'nt April 1st, is it ?
   I have two questions -how does the neural network 'know' that the purpose of the simulator is to keep the plane flying?
   Question 2 is :How long before someone other than me, mentions the word 'Emergent'?
   Wheres the rabbit?

Comment by Mike Godfrey — June 4, 2007 @ 1:41 pm

7. keiths Says:
   June 4th, 2007 at 2:47 pm

   Mike Godfrey asks:

   I have two questions -how does the neural network 'know' that the purpose of the simulator is to keep the plane flying?

   Hi Mike,

   The neural network doesn't 'know' that it should keep the plane flying, or even that it is connected to a flight simulator at all. It just receives a pattern of inputs from the electrode array and responds with its own pattern of outputs. It initially responds more or less randomly to all input combinations and doesn't succeed in flying at all. But the computer gradually trains it to fly by 'reinforcing' it when it performs correctly (i.e. if it pulls back on the stick when the plane starts to descend).

   What the network is learning to do is to produce the correct outputs for a given set of inputs, without knowing what either the inputs or the outputs represent. Unbeknownst to it, the inputs represent the attitude of the plane and the outputs represent movements of the control stick.

   The training is done by giving the network additional stimulation when it does the right thing. When electrical stimulation is applied, the active synapses will tend to strengthen, while the inactive ones will not. That makes the network more likely to produce the same pattern of activity the next time it receives similar inputs.

   Question 2 is :How long before someone other than me, mentions the word 'Emergent'?

   Demarse (the researcher) already did. It's the perfect word to describe a function arising from a network of neurons. He wrote:

   There's a lot of data out there that will tell you that the computation that's going on here isn't based on just one neuron. The computational property is actually an emergent property of hundreds or thousands of neurons cooperating to produce the amazing processing power of the brain.

Comment by keiths — June 4, 2007 @ 2:47 pm

8. Mike Godfrey Says:
   June 4th, 2007 at 3:22 pm

   Hi Kieths,
   thanks for your reply.
   I realise that the neurones respond to a pattern of stimualtion,I guess my question is; how do you train a neural network to do anything?
   Behaviour modifaction is out -so hows it done ?
   Thanks ,
   Mike

Comment by Mike Godfrey — June 4, 2007 @ 3:22 pm

9. keiths Says:
   June 4th, 2007 at 4:03 pm

   my question is; how do you train a neural network to do anything?

   Mike,

   You skipped over the part where I explained that:

   The training is done by giving the network additional stimulation when it does the right thing. When electrical stimulation is applied, the active synapses will tend to strengthen, while the inactive ones will not. That makes the network more likely to produce the same pattern of activity the next time it receives similar inputs.

   The key is that the additional stimulation is only applied in response to correct behavior.

Comment by keiths — June 4, 2007 @ 4:03 pm

10. Mike Godfrey Says:
    June 4th, 2007 at 5:49 pm

    Ah yes I remember it now!
    Cheers Kieths:mrgreen:

Comment by Mike Godfrey — June 4, 2007 @ 5:49 pm

11. kenter Says:
    June 7th, 2007 at 10:20 pm

    Other than the "Wow! Is this for real" factor, perhaps my suspicion that this is an April 1st hoax is only because of the gutter-talk so prevalent in my culture …

    "Dr. Duhm-ahss"!

    Ouch.

Comment by kenter — June 7, 2007 @ 10:20 pm

12. Thought Provoker Says:
    June 9th, 2007 at 6:58 pm

    Hi All,

    Here is a link to an experiment attempting to confirm the non-locality aspect of quantum mechanics using cultured neural networks…

    In recent times the interest for quantum models of brain activity has rapidly grown. The Penrose-Hameroff model assumes that microtubules inside neurons are responsible for quantum computation inside brain. Several experiments seem to indicate that EPR-like correlations are possible at the biological level. In the past year , a very intensive experimental work about this subject has been done at DiBit Labs in Milan, Italy by our research group. Our experimental set-up is made by two separated and completely shielded basins where two parts of a common human DNA neuronal culture are monitored by EEG. Our main experimental result is that, under stimulation of one culture by means of a 630 nm laser beam at 300 ms, the cross-correlation between the two cultures grows up at maximum levels. Despite at this level of understanding it is impossible to tell if the origin of this non-locality is a genuine quantum effect, our experimental data seem to strongly suggest that biological systems present non-local properties not explainable by classical models.

    EPR = "Einstein"“Podolsky"“Rosen" which is an aspect of "quantum weirdness". Einstein et al, argued it as a paradox, Bohm, Bell, Penrose, et al are embracing it as reality. Experiments with neural networks are providing support for this interpretation.

    For those just tuning in to this thought quest and it's relationship to ID…

    Based on Penrose's work, I am suggesting that everything is interconnected via quantum mechanics. Therefore, there is no such thing as "natural" random effects. Penrose-Hameroff model suggests consciousness is an artifact of quantum mechanic effects. My interest has earned me the designation of "Quantum Quack" in PZ Myer's Blog, Pharyngula.

    Dr. Dembski's description and argument for Specified Complexity is based on eliminating all (or most) natural chance hypotheses. Well, it turns out that maybe there is no such thing as naturally occurring chance. Just non-deterministic, non-algorithmic and non-local quantum effects that show an ability to demonstrate retroscausality.

    Provoking Thought

Comment by Thought Provoker — June 9, 2007 @ 6:58 pm

Leave a Reply

You must be logged in to post a comment.