On Sun, Jul 10, 2005 at 11:49:53PM +1000, Stathis Papaioannou wrote:

> >3) Combining General and Particular Architectures
> >Fusing information to combine apriori knowledge of general architecture 
> >brain functions, and particular architecture data obtained from in situ 
> >functional measurements (e.g. fMRI), neurological and psychological 
> >measurements, as well as self-analysis, it may be possible to reconstruct 
> >a functional copy of the brain close enough as to be indinstinguishable 
> >from the original by the owner. How does the owner knows it is 
> >indistinguishable? This is a whole topic. He could for example do a series 
> >of  partial substitutions to find out if it feels the same or not. For 
> >example, he could substitute in sequence the visual cortex, the auditory 
> >cortex, some of the motor functions....
> >
> >We may be closer to this goal than you think.
> OK, I agree it is possible, and I'm glad nobody is insisting that just the 
> arrangement of neurons and their connections, such as could in theory have 
> been determined by a 19th century histologist, is enough information to 

Exactly; it's a strawman position. Nobody claims a 5 m resolution satellite
photo shows you what brands of pizza that shop on the corner is selling.

> emulate a brain. I think we would need to have scanning resolution close to 
> the atomic level, and very detailed modelling of the behaviour of cellular 
> subsystems and components down to the same level. I don't know how long it 

You need this level of detail only initially, to obtain empiric system
parameters for an abstracted system level. You might want to reach down to ab
initio level of theory, to obtain missing parameters for an MD simulation, to 
switching behaviour of an ion channel, depending on modification, to obtain
computational behaviour of a piece of dendrite (of course, you can also obtain
that empirically from e.g. voltage-sensitive dye/patch-clamping). Even then,
the actual simulation unit could be a few layers up, at abstract neocortex
columns, or similiar.

In the end, you have to destructively scan an animal to obtain your very
large set of numbers, to enter into your simulation. Transiently, that 
might involve sampling some voxels at a high level of resolution, very
possibly submolecular. That level of detail might be present in the voxel
buffer, transiently, before being processed by algorithms, and destilled into
a much smaller set of small integers.

> would take to achieve this, but I know that we are nowhere near it now. For 
> example, consider our understanding of schizophrenia, an illness which 

If we had fully functional (discrete, fully introspective, traceable)
models of individuals having schizophrenia, and controls, finding structural
and functional deficits resulting in the phenotype would be effectively

> drastically changes almost every aspect of cognition. For half a century we 
> have had drugs which ameliorate the psychotic symptoms patients with this 
> illness experience, and we have been able to determine which receptors 
> these drugs target. But despite decades of research, we still have no idea 
> what the basic defect in schizophrenia is, how the drugs work, or any 

We don't have methods with sufficient resolution, that's all.

> clinically useful investigation which helps with diagnosis. Although fMRI 
> and PET scans can show differences in cerebral blood flow compared to 

fMRI has voxel sizes at several mm^3, and temporal resolution of seconds. MRI
microscopy does much better, but only works on insect/mouse-sized samples.
Nondestructive methods do not scale into the volume.

> control subjects, this is a secondary effect. The brains of schizophrenia 
> sufferers, looked at with any tools available to us, are essentially the 
> same as normal brains. In other words, a very subtle, at present 
> undetectable, change in the brains of these patients can cause gross 
> cognitive and behavioural changes.


would seem to disagree. 

Eugen* Leitl <a href="http://leitl.org";>leitl</a>
ICBM: 48.07100, 11.36820            http://www.leitl.org
8B29F6BE: 099D 78BA 2FD3 B014 B08A  7779 75B0 2443 8B29 F6BE

Attachment: signature.asc
Description: Digital signature

Reply via email to