Known Knowns, Known Unknowns and Unknown Unknowns
Mr. Rumsfeld’s Famous Quotation from 2002
Cybernetics tells us that “AI” can never be controlled. Stewardship (“oikomonia”) can be taken of the people and technology; Guidance (“kubernetes”) using “controls” can be applied but “control” can never be established.
The “known unknowns” are comprised of a set of assumptions derived from the evaluation of the system that maybe useful in most cases (indeed may reduce the cost of resources required to support the system) but potentially pathological to the system in the case of exception. There cannot be a feedback mechanism to handle assumptions.
The “unknown unknowns” are by definition beyond control.
The “AI” process engine cannot be controlled as there will always be a cybernetic “open-loop”. This is because:
1.Known unknowns – the behaviour of the engine cannot be predicted.
2.Unknown unknowns – the range of behaviours cannot be reasonably understood therefore constrained.
Therefore, a “boundary” must be set for its actions i.e. the system must be closed, if necessary, by the physical disconnection of a given “AI” engine from its engagements.
Both internal and external systems to the entity must be constrained. Cybersecurity is one mechanism for the constraint of “AI” where a boundary for its action has not been set either accidentally or deliberately.
The question is whether a sandbox, strait-jacket, padded cell or quarantine station is to be used.
Leave a Reply