AGI Predictability - I will be using this to refer to the Predictability obsrved by a User observing or Interacting with the AI Program.
AGI Predictable - AI Program results / actions that are overall Understandable or Expected to an observing User. Often there may be Multiple different Actions or Behaviors that, any of which, may still be considered 'correct'. Also, these results may vary or deviate in Behavior (change over later repeated actions).
Mathematically Predictable - most often, mathematical equations are expected to always generate the same output answers for specific input data. So, Mathematically Predictable refers to having completely expected output data for specific input data. Always Right / Wrong Answers. Note: this differs from the AGI Predictable term described above.
AGI Unpredictable - would be an AI Program that reacts with generally Not Understandable Words or Actions to an observing User.
Ambiguity - refers to deviations in behavior, but which is still generally understandable - so correct behavior. This includes deviations (result differences) in repetitive situations.
Ie: Input of Device Overload, could yiled Results of:
- 1) Triggering User 'Warning' Alarm,
- 2) Adjustment of Lowering Power to Device,
- 3) Shutting Down of the Device,
- 4) Shutting Downof Entire System.
Any, Most or All of these Results would probably be considered 'Correct', but are all different - making the Answer / Behavior 'Ambiguous'. Later behavior may switch after Learning Results (say from Shutdown to Warning , or Warning to Shutdown - depending upon actual Results observed / Learned - or User suggestions).
Learn - ability of a Mind to use Observervatons of Info and Situations and Actions - to modify its own Memories, Thoughts, and Ideas - to Plan and Change Future Thoughts, Ideas, Decisions, and Actions.
Imagine / Imagination - to Model a World Observation in the AI Mind , or to Model Idea(s) in the Mind including Changing and Experimenting with the Idea.
Creativity - describes the deviations (and often the result deviations / differences between redundant actions). And at the same time, giving the capability to Change and Generate New and Unique Ideas.
Creativity often results in high Ambiguity, and may Lower Predictability some in the complexity of behavors.
For example: the AI program may not just answer with a 'yes' - it may also use 'ok' or 'yeah' in later questions / situations.
These are obviously synonomous in results , but may seem different to the user. Clearly there is nothing wrong in this behavior. Obviously, this is a simple example.
More complex situations may show 'Unpredictable' Actions - but still Generate Acurate Answers / Results. Again, the Means should tend to be proven by the End Results.
I want to express this corelation between Creativity and Ambiguity and Predictability, considering that more complex Behaviors (in Psych and so complex AGI) often lead to Ambiguities in their Behavior.
Independence - the behavior that the AI Program acts in its own in Alternate Behaviors that goes beyond what the User expects or wants. This should not be considered negative, as long as the Behavior does not interrupt activities, distract from the expected goal, nor 'abruptly changes the subject'.
Autonomous - related to Independence. Mind that Independently decides and executes its own Actions and Plans.
Understand / Understanding - ability to Observe & Modelize World and/or Ideas.
Consciousness - the ability that the AI Program can: Process Input data to make short and longterm decisions, Generate New Ideas from Observed Info or Situations, Solve Problems, Imagine World Data and New Ideas, Learn Info and Actions and Behaviors, and Plan for its own future and benefeit.
In short, Consciousness - means to Observe & Understand & Correlate & Modify & Solve Ideas and Situations - for its Immediate circumstances and Future.
My Goal is to introduce an Autonomous Intelligent & Learning AI.
With AGI attempting to model the Human Mind, I do not see the Mathematical Predictability. I think few people appreciate being told that they are 'Predictable' - so why should we expect AI to be Mathematically Predictable either ?