Frighteningly, as if we didn't yet know it, the age of the Cyborg is here: namely, the insertion and adhesion of electronic equipment which will maintain long-term functional contact between nerve cell networks and microelectrode arrays.  In theory and practice, this has been around for awhile, such as in cochlear implants, but the signal has not actually been the same as we hear it without, that takes time to adjust to, as it is a sensory system prosthesis, as it would be for retinal and visual implants.

Given that sensation or perception mechanically ends at the receptor, namely retina or tympanum, and thereafter is signal, the task is just signal to noise ratio balancing after that. The extension of neuronal prostheses to the central nervous system is just a few moves in understanding the computational and cognitive properties of the brain and related nervous systems.

So the book represents a historical and scientific hierarchy of what has been done, what is now possible, and what will follow in the future.

The 16 chapter titles run along this developmental continuum: We have made the deaf hear, now what? Microelectronic array for stimulation of large retinal tissue areas. Imaging two-dimensional neural activity patterns in the cat visual cortex using a multielectrode array. And that is only the sensory systems. In terms of neural representations, the titles continue: Brain parts on multiple scales: examples from the auditory system.  A protocol for reading the mind. Cognitive processes in replacement brain parts: a code for all reasons.  Mathematical modeling as a basic tool for neuromimetic circuits.  Real-time spatiotemporal databases to support human motor skills.  And now to interface this with real tissue: Long term functional contact between nerve cell networks and microelectrode arrays.  Building minimalistic hybrid neuroelectric devices.  The biotic/abiotic interface: achievements and foreseeable futures. So now to the hardware: Brain-implantable biomimetic electronics as a neural prosthesis for hippocampal memory function.  Brain circuit implementation: high-precision computation from low-precision components. Hybrid electronic/photic multichip modules for vision and neural prosthetic applications.  Reconfigurable processors for neural prostheses.  And lastly, The coming revolution: The merging of a computational neural science and semiconductor engineering.

That one paragraph above is enough to wreck Microsoft Word's spell check forever. A whole new, emotionless world of implants is growing and seems feasible without any doubt, if the mathematic models are right, as they should be, as Sony has just unveiled a robot with human-like movements.

In August 1999, the contributors to this book met in Washington D.C. and this book flows from that, showing to the world that it will be possible to mathematically model the functional properties of different regions and subregions in the brain, design and fabricate microchips incorporating these models, and create tissue-electronic interfaces to integrate the hardware and the brain functions they subserve. Bidirectional communication would then allow the brain to restore sensory, motor or cognitive functions lost through damage or disease (preface vii).

Two chapters deal with the issue of dealing with the dynamism across time of presenting neural representations, and one with the issue of dynamism related to ongoing experience of novel representations.  Memory systems are most complex in their ongoing mercurial nature, dealt with in two chapters which emphasize the difficulty, and necessary strategies for dealing with heteromodal input into the limbic and hippocampal system particularly, as well as then provide processed output into complex neuronal subsystems in the global information encoding and processing system of the diencephalon.  In this latter regard, one chapter focuses on the difficulty of modeling the hierarchical nature of information processing systems, and another tackles the binding of events across time to facilitate sensorimotor coupling, no easy mathematical modeling task.  Despite these difficulties, hippocampal and thalamocortical systems' modeling has given rise to very large scale integration techniques for direct and parallel streaming of information, dealt with in two separate chapters.  The problems of high density connectivity that arise, as well as those in multiple pattern storage for context dependent connections and functions are also dealt with in separate chapters.  Drawing on these last chapters, the final chapter looks into the present and extrapolates into the future of next generation computational engines.

IBM have made it clear that their engineering and technology services are using Blue Jean Computers and Cell Broadband Engines to enable just that in the future, and that such computational engines are a chip-throw away.

As Howard Eichenbaum muses:

Wouldn't it be wonderful if we could develop a means for interpreting the patients' recollections and plans so that artificial speech and other robotic devices could express his memories and intentions?….In my view, we now have sufficient data on the anatomy and physiology of the brain to encourage optimism that such a device could be developed (page 91). 

I hope he is talking about a living human being….but he is talking about mind reading, and his title makes this explicit.  Hampton and colleagues make explicit that stem cells and growth are not on their agenda, but replacement by a synthetic component is (page 111).  They of course have to deal with the way the brain filters information, with the final functional outcome different somewhat to the entire 'ensemble' in their terminology. Just to illustrate, in order to deal with this problem of ensemble versus functional outcome, they have to deal with say, the CA fields of the brain.  These have input which involves the CA3 and Mossy Fibres, the CA1 and both Schaeffer Collaterals and the Subiculum.  Inputs arrive which involve serotonergic input from the Raphe Nucleii, the noradrenergic input of the Locus Coeruleus, cholinergic input from the Nucleus Basalis of Meynert, inhibitory input from the cerebellar vermis, amygdala and anterior cingulate, as well as insula and frontal orbitobasal input and so on: information processing is a second order machine, but it doesn't seem to faze these research engineers as they produce mathematic models to deal with all of this, as they glibly note "Cognitive neural codes are dichotomies of referent information" (page 114) which can be decoded and programmed. That is not to say that replacement parts have to mimic or process information in exactly the same manner as the original circuits, its just that the functional codes that they generate have to be compatible with the ensembles they represent and are a component with the behavior or cognitive processes that they support. The question that arises is whether Gage will still be Gage, after treatment.

© 2007 Roy Sugarman

Roy Sugarman PhD, Director of Clinical and Neuropsychological Services, Brain Resource Company, Ultimo, Australia

Categories: Psychology