MyKell

Unsupported claim. They seem to be necessary but not sufficient!

Again, you are referring to nitric oxide (NO) and Carbon monoxide (CO) when talking about the gases. Ironically both knockouts of the enzymes that make them (nNOS and HO2) give phenotypes with normal memories

Again, the memory of a computer is "representational", ours has to be "non-representational". It is not coded in bits like a computer is. All the data from people with brain injuries point to that. So examples from neural networks won't do.

m00ner

We'll eventually find some way to be able to reincarnate people with their memories still intact.

MyKell

Sure! wonderful argument...

Keith Russell

Greetings:

DNA strands do not contain memories, only the building blocks for our physical structure.

Clones are not duplicates of the people from whom they are cloned. They are only physical duplicates; twins, not complete copies.

Keith.

m00ner

I know, that's why we would resurrect, not clone.

...Or maybe we could come up with a way to clone a person and insert memories.

The Insomniac

I think many of you in this thread are overstepping your bounds, WW1 and WW2 has occured, and WW3 will too without reinvention from the ground up. I see a short in this circuit.

<shrug>

No Sleep

"If humanity goes extinct, where does one place the fault?" -- The Insomniac

excreationist

Neural networks could also be implemented using electronics rather than software... BTW when it is in software, you're supposed to ignore how it is actually represented in bits on a serial computer and assume it is a neural network that runs in parallel (doing multiple things at the same time).
Could you explain what you mean by "representational" and "non-representational"?
What if there was a robot that had been trained to perform some task, that used a very complex neural network... would its memories that are within the neural network be "representational" or "non-representational"?

What do you mean about brain injuries? - my impression is that sometimes brain injuries are fatal, or they lead to massive losses of function that can often be learnt again, or no obvious difference in the person. The same happens on software-based neural networks - assuming you destroy the neural network weights rather than other parts of the software like the operating system. Neural networks in generally "gracefully degrade" and are "fault-tolerant" - but if you tampered with the OS on the computer then it would ruin things much faster. (i.e. it would crash/freeze)

MyKell

A representational memory means the coding of the outside world in some kind of representation that can be retrieved unchanged when wanted. This is the kind of memory humans have obviously generated in most of their artifacts. This memory, however, requires that the world be presented in an anumbiguous code. (just like a computer language, a thread of zero and ones). And the system having an error-correcting mechanism when faults occur while retrieval of a memory (computers and neural networks).
For a biological system like the brain, which also a selectional system rather than a constructional system, degeneracy is inevitable.
selectional : during development, neurons are not wired according to some code. There are much more neurons to begin with than finally left, so the ones that make successful connections survive.
degenracy: more than one structure (in this case a neural circuit) can do the same job.
Information to the brain is presented in a rather ambiguous way. Perceptual categorization is carried out by brain circuits wihtout a priori knowledge. There is no neural code. At least, there is no evidence for one.
Selectional mechanisms also work in memory. The very degenerative nature of the brain, makes it possible for many neuronal circuits to recategorize information whenever a memory is to be retrieved (i.e an action is to be repeated). That is why the retrieval of a memory is not simply copying the stimuls, it is active recatogrization of the original interpretation of the brain for that stimulus, which could occur in multiple ways.

Now for brain injuries, the thing is different. Memories seem to go fainter, but never a chunk of information (like for instance, very specific phone number, or a very specifc knowledge of places) is lost. What I mean by injury is any insult to the brain... mainly strokes, which could be nothing but very small microthrombi.

Cool?

Now I have not associated any of my previous statements on memories with an phenomenology. I don't think Neural Networks will ever be able to feel anything. Synapses are far away from qualia.... Not that phenomenology is restricted to biology, just that the current state of simulation is poor.

Excuse me for blabbing in the end, it's 6 in the morning over here

sweep

I think it would be more accurate to say that those which recieve the most input survive. But then this might only apply to motor neurons, if my memory serves me right.

Secondly neurons not wired according to code, but coding does affect structural changes, during stimulation. allow me to go into depth:

Throughout the nervous system, there are many neurons that are receptive to the neurotransmittor, Glutamate. On the post synaptic membrane of many neurons, there are two kinds of receptor, that are activated by glutamate. One is the AMPA receptor; this is ligand gated (meaning that the molecule glutamate activates it). The neurotransmitter glutamate binds to the receptor allowing the influx of calcium ions, and changing the voltage inside the post synaptic membrane. IF, by excitation, enough neurotransmitter is release, more AMPA gates are activated, thus changing the voltage inside the membrane (POST synapse: The synapse is the gap between neurones that prevents continual transmission) The second gate on the post synaptic membrane is the NMDA channel. This is both LIGAND and Voltage gated. glutamate attaches to the receptor, but magnesium blocks the channel preventing the flow of positively charged calcium ions. But when enough AMPA receptors are activated by glutamate, the change in potential with the post synaptic membrane changes enough to unblock the neighboring NMDA receptor allowing a much greater influx of calcium.

This is where 'Long term potentiation' comes into play (bliss & lomo, 1973). Tetanic stimulation of a motor nerve can produce a stable and enduring increase of the response of that nerve. In light of the above information a sudden influx of calcium ions has been shown to produce lasting changes within the structure of a neuron. This includes dendritic branching, a late development in the process of a neuron.

the following has been extracted from 'biopsychology (1999), Rosenweig, Lieman & Breedlove

"1. Strong stimulation of the neuron leads to a rapid increase in the intracellular concentration of calcuim ions.

2. Increased calcium ion concentration activates protein kinases (including calmodium kinase (CaM), PKA, PKC, etc.) which phosphorylate proteins.

3. Activated kinases bind to CREB (CAmp, resposive element binding protein), leading to production of the transcripts (mRNA) or immediate early genes (IEGs)

4. Many IEGs code for transcription factors. These proteins (produced in the endoplasmic reticulum) enter the nucleus and regulate the expression of particular late effector genes (LEGs)

5. Transcription of LEGs leads to syntheiss of proteins, including enzymes and structural proteins. Some of these proteins are necessary to induce LTP.

6. Many of the proteins sythesized are transported down the axon and dendrites to alter the response of the neuron to additional stimuli."

excreationist

There was a documentary I watched a while ago - I think it was by Susan Greenfield - perhaps it was "Brain Story" - and there was an old man who gradually lost the ability to classify pictures of animals. He would see a cat(?) and say its a dog or something (since he forgot about the word cat)... they said his memories (of categorizing things?) were being lost from the most recent going back to the earliest. Unfortunately I can't find any links about it at all. It seems counter-intuitive but I remember watching it (if that counts for anything).

If there are only small then it is likely that they wouldn't completely destroy information like phone numbers - the patterns for the phone number would be spread across many neurons. If only some of the neurons were affected then a lot of the number would be able to be recalled, and some parts might be fuzzy - e.g. you mightn't be sure if it was a "6" or a "9" - or if it was "52" or "25". (BTW, this applet I made demonstrates that idea, though it takes a while to explain)

I think neural networks are more about learning and predicting things - i.e. the source of intelligence - perhaps a control system at the inputs and outputs would be more directly related to qualia.

Did you stay about that late? I have been staying up to 2:30am or 3am sometimes... I've got to go to sleep early.