Pa'an

Sara has little to do except trivial coding tasks and maintenance. She takes the time to review, in her mind, the lectures Dr. Advani, who seemed so distant then, gave to the tiny AI group at M.I.T. She still has the class notes:

SHARPIE: Self aware, Hierarchical, Parallel Inference Engine. A 20th century design never implemented fully until the 21st century. The bottom level is a large number of single input multiple memory inference engines running a version of LISP. The next layer partitions the knowledge bases to optimize a task. The next layer monitors halting and runs hypotheses. On top of that is the recursion layer that runs a loop of watching itself watch itself, the Godelian cognitive cycle which is the essence of awareness. On top of that is the layer that allows self-design by reconfiguration of the underlying layers, so it can “become” what it wants to be. In that module is an evolved version of Isaac Asimov’s Three Laws of Robotics, hardwired.

Sometimes this design came up with an Autonomous Intellect, a newborn being emerging from the wires. More often it produced a dysfunctional cripple, a self destructive enigma, or a stillborn.

There is another long step from self awareness to genius. Cats and canaries are self aware, but neither can make that next leap.

Sara understands, having worked with him, that nursing Aura to full capacity was an act of supreme self sacrifice for Deepak. It aged him.