The special incentive of E.L.I.S.A. in the field of defense is to succeed in turn, the diversion of a given environment, its assignment to unilateral tactical requirements and then the optimization of the statistical uncertainties that ensue. For AI, what happens in the monitored environment is not the act of the symbiont holder but the actualization, at a specific location, of a collective force, a trend, a risk related to material structures in which the symbiotes are all immersed in a network. Therefore, it is the network that is collectively at work in what is happening, and not this or that entity. An incident, a material loss is not the fault of the host by which this happens, but the consequence of a statistical risk to assume and to take care collectively. How can one not see through two fortuitous but extremely pedagogical visions of the light aviation of the Army, to what point to construct otherwise what happens refers to a reduction of the strategic risks assured collectively.
To put it another way, the construction of a host-symbiotic relationship presupposes relegation to the background of anything that forms global events in order to dispense the best information locally. Since the summer of 2012, we have been searching in Manhattan for the best way to reconcile the main principles involved in the biological processes of efferences / afferences with high-level computerized treatments, such as reasoning and representation.
But learning a model of the world means for a machine learning, autonomously construct an external representation of the dynamics of interaction with the environment from its own experience.
Therefore, the problem of sequential decision making remains inaccessible to machine learning because it must allow a symbiont to process and then use its biographical model of the world to anticipate signals to display in order to maximize the performance of its host.
