“I’ve viewed the bridge imagery,” her soprano was almost as clinical, as detached, as before when she went on once more, “and aside from drawing Simon’s pulser and opening fire, he never moved from the instant whatever was controlling him took over. He simply stood in one place, held the trigger down, and swept his fire across the bridge. Looking at the security imagery of the Havenite ambassador’s chauffeur shooting Admiral Webster, you see very much the same thing. He draws his weapon, he opens fire—hitting three other people, not just the Admiral—and simply stands there until he’s shot in turn. No attempt to escape, no effort to take cover or evade return fire, nothing. We haven’t had the opportunity to look at any imagery your security people may have of Colonel Hofschulte, so I can’t say how consistent that is in his case, but that’s the pattern we’re apparently seeing.

“Now, according once again to my dad, there has to be another component, some sort of organic AI, you might say. His best guess is that it probably sets up residence in some corner of the target’s brain, but it wouldn’t necessarily have to be located there as long as it has access to the central nervous system. Presumably the AI is issued with a set of triggering criteria that it looks for before activating whatever ‘muscle memory’ may have been installed in the nanotech. Obviously, the criteria can’t be too complex.”

Rabenstrange nodded again. The ancient cybernetic hope of achieving true sentience in an “artificial intelligence” had never been realized. Enormous strides had been made in crafting “brilliant programs” which could mimic intelligence—hence the continued use of the term “AI,” even though it was technically incorrect—but those programs could react only to parameters the programmer could anticipate. The ability to discriminate triggering criteria in something as complex as normal human interactions was notoriously complicated, unless the programmer had a very specific grasp of the interactions likely to arise or could build in a mechanism the program could use to gain additional information and extrapolate from it. An AI designed to deal with customer service inquiries, for example, or to operate an air taxi, and which had the opportunity to clarify situations and desires by asking questions, could give an incredibly convincing imitation of genuine sentience within its area of competence. Outside that area of competence, however, and without that ability to expand its information base, the situation was completely different. And if Dr. Harrington was right, the AI in this case certainly wouldn’t have the opportunity to ask any “clarifying” questions before it acted!