From just-in-time to just-ahead-of-time
Anbar Province, Iraq (Summer 2006)—A U.S. Navy SEAL Task Unit was assigned one of the most difficult and dangerous missions imaginable: rescuing a teenaged Iraqi civilian being held for ransom by an Al Qaeda-linked terrorist group. With no time to lose, Unit Commander Leif Babin began briefing everyone involved in the operation on his plan for secretly approaching the building in which the hostage was believed to be held. The element of surprise was critical. But as they finished loading up their gear and started moving out, an intelligence officer burst into the room. They had just received word that the enemy had set up heavy defenses, including IEDs buried in the ground and machine guns bunkered inside the building.
What last-minute changes do you think the SEALs made to their plan in light of this new information? The answer: none. The reason: They had already anticipated the presence of such threats, even though none were indicated right up to the time of their departure. The result: Mission accomplished, with no friendly casualties.
As a KM’er you might ask: Will an AI system ever be able to automatically plan such high-risk operations that hold up under rapidly changing conditions? The answer is a resounding yes. Welcome to the world of anticipatory systems.
The co-evolution of human and artificial intelligence
Humans and, to some extent, all living things are anticipatory by nature. We survive not just by reacting, as in the “fight or flight” response, but also by anticipating, trying to stay one step ahead. Yet we still get blind-sided by 9/11, active shooters, financial market crashes, deadly virus outbreaks and other so-called “black swan” events.
It isn’t that such events aren’t anticipated. Rather, the warnings tend to be overshadowed by our biases, amplified by a growing dependency on data and logic (algorithms). Unfortunately, purely data-analytic approaches rarely take high-impact, low-probability outliers into account. Through our many decades of research into the knowledge sciences, we’ve determined that if AI is ever going to live up to its promises, it needs an architecture that integrates three levels of functionality: memory, awareness and anticipation.
The first level, memory, is where most of the evolution in computer technology has taken place. Because of our obsession with raw computational power, most AI architectures have been focused on bringing memory and processing capacity up to the level of the human brain.
The second level, machine awareness, attempts to give meaning to the data stored in memory. The emphasis is on what is known and how it is known. This has been evolving at a much slower pace and falls into the realm of ontology, epistemology and the semantic web.
The third level, anticipation, has barely been scratched. Throughout history, bio-mimicry has been the usual starting point for technological breakthroughs—from airplanes to robot arms to computer vision to natural language understanding. We study how natural systems work, then attempt to do the same using machines. Designing and building anticipatory systems will likely follow the same trajectory.