[Expectant_Philosopher (OP)]

Couples, team mates, fellow crew members know each other so well that their interactions can be governed by unspoken intent.  Can we program artificial intelligence to respond to intent?  In such a way a program interface will react almost as if they can read our minds.  The programming for this can occur in and take advantage of the time scale difference between man and machine. In the time scale differences of telecommunications and man we send video messages across the globe through the Internet, and everything looks smooth as long as the transmission delay is kept below 150 milliseconds.  In an intention engine we would just be speeding up the reaction time of the AI device, by giving the device the ability to sense cues of voice tone, body language, and phrasing, then preloading probable possible outcomes.  Then when the estimated cues intersect with real inputs, the AI computer will react as if it already knew what you wanted.  The incredible processing speed just makes the reaction seem smooth, although there is still a small delay.  It would make the AI responses seem on par with our human partners.

[alancalverd]

Why not? As long as you have a consistent and measurable cue, you can infer intent. Predictive text is a simple example.

But the applications are limited and must be noncritical. Talking to dogs and pilots often involves a "hold", then an instruction, followed by a trigger word. "Wait, fetch slippers, go" works for some dogs, and a departure instruction usually begins with "Hold position. On release, track x for y miles then.......blah blah, numbers, numbers...." "Alfa Charlie copy blah...." "Readback correct, stand by....Alfa Charlie cleared for take off" - all hell breaks loose if you infer intent, or don't confirm the data, before release!