Investigating what occurs when AI systems with different objectives interact raises questions about coordination, competition, and conflict resolution among AI systems.
If you mean "different opinions about the same goal", you might launch a discussion. If you really mean different goals, the answer is war.
Gilbert Brands suppose if 2 or more goals reside in the same meta application? Suppose this meta application has paradoxical goals? Suppose I build something AI that needs anonymity but I need to weed out the trolls, predators, and other unsavory types, how would one reconcile these paradoxical positions?
In order to achieve a goal, rules are needed to calculate the individual steps. These rules require a clean axiomatic basis to be consistent. Starting from a specific situation, the different goals must be compatible in terms of the rules, i.e. be achievable at the same time. Whether a way can then be found is another question.
If there are different axiomatic bases (e.g. mathematics and religion), reconciliations can be found, which then probably favor one goal if they are not compatible.
If everything is based on circular superaxioms (e.g. climate change changes everything - it's raining today - that's climate change), then some chaotic behavior is likely to result if the goals are not compatible. The same may arise if you have no axiomatic basis at all.
Of course, you can also try to teach the system to have ideas. Ideas, in turn, may be rules, but this has to be tested in axiomatics. You either have something new or throw it away.
Lutsenko E.V., Golovin N.S. The revolution of the beginning of the XXI century in artificial intelligence: deep mechanisms and prospects // February 2024, DOI: 10.13140/RG.2.2.17056.56321, License CC BY 4.0, https://www.researchgate.net/publication/378138050
There are many varieties of paraconsistent logic that can deal with or circumvent inconsistencies and conflicts in one way or another. Perhaps one or more of these systems of logic might be useful for the design and implementation of AI.
https://iep.utm.edu/para-log/
https://www.jstor.org/stable/44084588
Conference Paper Paraconsistent Annotated Logic Artificial Intelligence Study...
Article Neural Network Explainable AI Based on Paraconsistent Analys...
Karl Pfeifer does that make paradoxical logic absurd (in contrast to paraconsistent logic)?
Tieu-Tieu Le Phung To the limited extent that I'm familiar with the expression, paradoxical logic is mostly nonformal or semiformal, i.e. it is articulated mostly in natural language. So I would say that paradoxical logic can be mapped onto a formal system of paraconsistent logic to be used by an AI in automated reasoning. Of course a lot of trial & error and tinkering on both sides would be involved in initial setup and implementation. Also, a mapping and subsequent dry-run processing might reveal some inherent absurdities in one's particular paradoxical logic.
In my opinion, the outcome of the interaction of two systems equipped with generative artificial intelligence technologies with different goals depends on how the algorithms responsible for the discussion were developed, whether a real-time interactive process of analyzing another opinion, other goals, arguments in support of a certain other thesis, etc., was included. In addition to that, whether tolerance of the discussant, another opinion, thesis, the possibility of reaching a consensus, developing a common solution, which may be a mediation against the previous two other claims, sentences, theses,, goals, etc. on the same topic, has been taken into account.
Regards,
Dariusz Prokopowicz
- Badges
- Science topic
- Similar topics
- Psychology