Expert Systems

Intelligence vs Expert System

The difference between a true Artificial Intelligence and a mere Expert System is the ability to think. A true AI has this ability, an Expert System does not.

When an Artificial Intelligence encounters something that is outside its programmed parameters while at first usually disoriented it is able to react after working out what it considers the proper response. This gives them the ability to initiate spontaneous action similar to natural intelligences. Intelligences, whether artificial or natural, roll their Intelligence+Interface+Bonuses for Interface.

When an Expert System encounters something outside its programmed parameters it does not respond. This is because it lacks the ability to think: it merely follows a list of conditions that trigger responses. These conditions and responses can be made incredibly complex, making it sometimes difficult to tell apart a complex Expert System from a true Artificial Intelligence. Ultimately even such complex Expert Systems have their limits: they can not transcend beyond that which their creators have given them. Expert Systems lack the ability to initiate spontaneous action, although they can be made to react to any event the original creator was smart enough to anticipate. Expert Systems roll the program’s Expert System x5 up to the system rank + the program’s interface bonus for Interface. The program’s interface bonus is considered to be the skill of the expert system. This might be needed for confirming critical success and failure, or if the expert system operates machinery in meat space.

Expert System Levels

The expert system function is special, since it is never executed. The expert system function describes the complexity of the program’s internal logic. Because there is considerable variation within expert systems, levels are used to describe just how complex the logic is.

In game terms the expert system function executes another function within the same program when some predetermined condition is fulfilled. Complex expert systems are difficult to code: each level of expert system beyond 4 adds 1 to the base TD of a program. This is in addition to the TD increase from an increasing number of functions.

The term “well defined environment” means an environment in which everything is exact and easily measured. Virtual reality is a good example of such environment. A fuzzy environment is inexact and difficult to measure. The real world is a good example of such environment. Human social interaction is even fuzzier than reality.

Level Just how smart is it? Example applications
Level 1 The expert system is capable of initiating a single function based on a single outside trigger. This reaction-trigger pair permits the expert system minimal autonomous action. A firewall which executes the Stop (Gain Access) function when the Gain Access function is executed. A toaster which toasts bread for 4 minutes when the button is pressed.
Level 2 The expert system has multiple trigger-reaction pairs similar to a level 1 expert system. Systems at this level are usually capable of simple autonomous operation in a well defined environment. A scout program which systematically attempts to enter all rooms in a system, executing Gain Access when it encounters doors it does not have credentials for. An industrial robot which builds a car from component parts according to specifications and stops if it detects unrelated movement nearby (to prevent accidents).
Level 3 The expert system has a large number trigger-reaction pairs similar to a level 1 expert system. As long as the environment is well defined the expert system is capable of complicated autonomous action. Needs an example program. A building capable of receiving voice commands, controlling the climate, managing simple household robots, calling the cops if its windows or doors are broken, or the fire brigade if it gets too hot.
Level 4 An expert system at this level is complex enough to successfully carry out extensive autonomous operation in well defined environments. The expert system will attempt to complete its given task with the tools at its disposal. In game terms the player (not the character) who gave the expert system the task at hand can freely guide the expert system in its attempt to complete the given task. Cheap tech, medical and security scanners are often at this level. An agent program which is timed to make a run while the runner is in a prominent location. A police robot that can identify which persons are currently taking hostile action towards it and which ones are not.
Level 8 Expert systems at this level are capable of human-like operation in the real world. Attempts to imitate human social interaction are convincing enough to fool those without experience with expert systems, but nerds quickly realize when they call tech support that the sexy voice actually belongs to a robot. High-grade tech, medical or security scanners are usually at this level. A program with a good probability of success providing tech support. Needs an example robot.
Level 16 Expert systems at this level are capable of imitating human social interaction. Even experts can have a lengthy interaction with an expert system at this level without realizing the other end is an expert system. A virtual secretary capable of simultaneously serving everyone in a large corporate skyscraper.
Level 32 It is highly improbable anyone has ever coded an expert system of this complexity. Who knows what an expert system of this level would be capable of?

Expert Systems

Edgerunner Kohme Sevain