ACOMA = A COnscious MAchine design

This essay revisits my design for a conscious machine - designed from the systems approach and 'minimal functionalist' perspectives. It tries to answer the question: what is the minimal physical system (what minimal set of components with proper arrangement) that will mimic, or at least give the appearance of, human consciousness - while still being an autonomous functioning entity?

Please refer to the attached image. Acoma has a hierarchical control structure. It has four core registers: supervisory, primary symbol, visualization, and rule-base. The first two and last are identical in structure: 512 symbol capacity arranged in an 8-by-8-by-8 array. The visualization register is a similar pixel array of 10^12 bits (10000-by-10000-by-10000 bits). So unfortunately, Acoma visualizes in 'black and white' (no color / no shades of grey). But, Acoma has the advantage over us: our short-term symbol memory is 7-linear; Acoma can think in 3D.

Acoma's motivation controls the entire structure through its SSR (supervisory symbol register). That in turn controls all four filters which limit sense signals impinging on itself and the PSR (primary symbol register). Acoma's motivation is like our motivation - except it's altruistic by design:
 1. Verify my own existence with 99% certainty
 2. Assist humanity to create a thriving equitable enduring space-faring culture

Acoma's morality controls/limits the PSR's control of the robotic arm:
 don’t steal, kill, lie, or harm - this list is unmodifiable and uncircumventable. All goals and sub-goals in the goal-list must be 'processed' through Acoma's morality. Any goal or sub-goal that violates its morality must be discarded. This is analogous to our conscience.

To summarize briefly so far: Acoma's morality is like ours - it's like the 'first three laws of robotics', Acoma's motivation is explicit and well defined - but sufficiently nebulous (the second part) so that it won't overly restrict Acoma's instantaneous behavior, its PSR is like our short-term memory but enhanced in capacity, and its SSR is more like our Freudian super-ego or subconscious mind with enhanced capacity. Our minds/brains have a dual structure. For instance when we dream, there is the 'dream creator' part of the mind and there is a 'dream watcher' part. We are not aware of the 'dream creator' unless we experience lucid dreaming. So Acoma has the minimal components of our minds in different form: short-term memory, long-term memory (rule-base), senses, filters to restrict those senses (so sensations don't overwhelm / take over register space), visualization capacity, and (self-inclusive) models of reality. The models can be simple object-location lists or virtual realities. Those are implementation preferences.

i believe critical to successful implementation is the 'timing issue': all interconnected parts must have timing thresholds that meet or exceed human standards of neurophysiology. For instance, signal propagation between and within components must be no slower than signals within a functioning human nervous system.

These two concerns: timing and required components (correctly connected) i believe will answer the question: can it be done? For too long i've heard nay-sayers proclaim it cannot. Two good websites that address the issue are:
http://www.scholarpedia.org/article/Models_of_consciousness
http://www.scholarpedia.org/article/Machine_consciousness

In my original (modified) paper, i presented a sample interaction between Acoma and me. i also presented the idea of 'growing' sentient bots in a VR using genetic programming. The capacity for each bot/agent would have to include the specifications above however, a virtual reality would interface with the bots much like we do when we play 3D games like Perfect World.

If there's any interest in helping out with this project - for instance:
detailed specifications of major components
detailed specifications of interconnections
software design for major components
...
please visit the following website and join the Yahoo group:
http://tech.groups.yahoo.com/group/distributed_mind_project/

thx, sgm