[ExI] Meditations on the million year mind.

[Alan Grimes]

om

Tonite I feel like mumbling incoherently about how you would go about 
engineering a mind to last a million years. The problem is much less 
trivial than it would first appear because the main issue is to maintain 
a youthful learning capacity even when memories spanning thousands of 
years are already stored. Other aspects of cognitive performance must 
not degrade eather. Any non-zero rate of degradation will turn out fatal 
over protracted time. This posting will make some brief comments about 
the physical requirements of such a mind but mainly focus on the 
architecture of the mind itself. Some consideration will also be made 
with regards to changing workloads and the inherent need to move beyond 
the baseline in terms of capacity and capabilities.

om

A cursory glance at the world will inform you that minds exist, 
primarily, in order to solve the problems of maintaining and operating 
their embodyment. These problems are protection, maintainance, and 
socialization. Additionally, if the embodyment is mobile, then it also 
involves navigation. Virtual 'embodyment' is frequently proposed so 
therefore I will address it. While there is certainly some value in 
being able to use a VR environment, it cannot serve as a primary 
environment because there will always be a physical substrate and 
because. By definition, any action in VR is utterly inconsequential on 
the physical layer and is therefore completely uninteresting. 
Furthermore, a VR intended to preserve a mind over long periods of time 
is, by definition, hell because it would cantinually have to force the 
mind to solve somewhat difficult but completely contrived problems just 
to provide the stimulation required to keep a mind in working order.

Regardless of what you wish to deem the embodyment to be, the mind will 
exist within a physical substrate. If there is one true thing about the 
universe it is that it loves to kick the shit out of substrates. Any 
feeling of safety you might have is both temporary and illusory. So you 
will need to engineer a countermeasure for every conceivable scenario 
that might damage your brain. But more importantly, you need a way to 
service and repair it to factory spec. So therefore it will not be a 
solid block of computronium, A more plausible configuration is .1 mm 
cubes of computronium packed in a classical organic phospholipid 
membrane or in a more trendy nanosystem. In any event it must be sturdy 
enough to do anything from play football to get mildly zapped by the B+ 
on a classic tube amplifier.

While the substrate is much fetishized, in the bigger picture it is 
profoundly irrelevant. The real issue is how the brain actually works. 
In the crudest sense, the brain works kind of like a flash memory. 
Because synapses can't form in specific locations, they must form in a 
shotgun fashion and then get pruned back. Similarly, a flash memory uses 
electrical currents to reset a bank of fuses and then selectively blows 
the fuses to write information. To erase information, the entire bank 
must be cleared. So as you are learning, you are progressively loosing 
synapses, as they are pruned back. There are some exceptions to this but 
that is the basic process.

Inevitably, however, the brain exhausts it's plasticity and simply 
cannot form new memories. Any approach you would use to grow new 
synapses, or add additional cortical columns will have the effect of 
erasing existing memories, or, if you must, the existing personality.

So therefore we need to go back to the drawing board and come up with a 
new system for thinking that can be expanded or re-configured as needed 
without unobtanium. Reconfigurability is important because there are any 
number of reasons you might alter your embodyment and your mind must be 
able to adapt as well. While we are at it, there are several crucial 
areas of performance that should also be addressed. These include the 
ability to deal with many tasks concurrently, as well as the depth and 
breadth of knowledge that can be brought to bear on any given problem, 
not to mention the capacity for higher-dimensional thinking, a deeper 
short-term memory stack, etc...

The first thing that needs to happen is the abstraction from the unit of 
data from the unit of computation. Doing this will open up the whole 
universe of data manipulation techniques. The current cortical column is 
both data and processor approach might be fast but it is not flexible or 
scalable.

What we find in the brain is a network of interconnected sub-modules, 
each with a relatively discreet function. A brain of the type I propose 
would start out with a hanful of functional modules dictated by the 
embodyment and what is essential to create a functioning mind. The rest 
would be created dynamically and without limit. The learning rate could 
be extrordinary by human standards and, on account of the ability to 
create new modules to process new ideas, could *OUT CLASS* any previous 
human genius (or upload thereof).

Because the system would be based on database theory, the amount of 
information could grow to the limits of the hardware, instead of being 
bounded by the limitations of the original equipment due to emulation, 
and because the original brain lacks capacity for expansion.

The next problem, is how to take advantage of such a system personally. 
The unchallenged standard proposal is to scan the existing brain 
(subsequent to sacrificing the patient in an occultic ritual involving 
removing the patient's heart and showing it to him while it's still 
beating; yes it's that primitive!).  But the prospects of extracting 
anything useful much less extracting the data from the processors in 
that secenario is extremely dubious. Actually, that's much too polite: 
It ain't going to happen. You just aren't going to extract enough 
information to dump it into the new mind architecture and have anything 
even remotely like the original.

Now, what Just Might Work (tm) is mind coalescence. You use a BCI to 
mind-meld with the new platform. The new platform will acquire your 
knowledge not by trying to un-tangle your rat's nest of dendrites but, 
*ghasp*, *shudder*, etc etc, by communicating with the living system. 
The next step would involve stimulating each cortical column to 
play-back all of it's information which can be remembered by the new 
system.

The ultimate question is whether the universe will allow you to choose 
your point of view. This is the most profoundly non-trivial question 
there is. Unfortunately, only experimentation will even begin to shed 
light on it. =\  Will your original mind ever become "spare" and 
disposable to your new mind? In any event, the best result will almost 
certainly be obtained by operating both minds in their joined state for 
as long as your original brain can be maintaned. Therefore it is urgent 
to construct the new mind at the earliest possible date, preferably ten 
years ago when I came up with this idea. =P

I was working on a story involving it, but then my DOS partition got 
wiped out so I had to manually hunt down fragments of the manuscript 
with a disk editor and re-construct the thing... FUN. Wanna read it? 
(350k file).

So, you can either spend the rest of your lives debating whether 
cryonics is better than plastination, and then, uppon your death, 
someone else decides to creamate you... Or you could develop AI, neural 
interfaces, and mind coalescence now and start the journy toward your 
first-millionth birthday... The choice is yours, punks!

-- 
NOTICE: NEW E-MAIL ADDRESS, SEE ABOVE

Powers are not rights.