Artificial Intelligence System Assessment

We received an urgent call from an international defence contractor with an immediate need to assess several AI solutions as potential components in a much larger system development. Exhaustive testing in the laboratory (and in the field) was precluded by the combinatorial complexity posed by thousands of fixed and mobile assets resulting in outputs to hundreds of networked nodes.

So our challenge was to come up with a means of comparison based upon the overall ability of each system.
To this end we derived an entropic measure of comparative intelligence based upon the number of sensors/inputs, actuators/outputs, processing power and memory. This was expressed as a relatively simple logarithmic formulation.

Somewhere