Unknown unknowns
The project aims to have a fighting chance against rogue artificial intelligence.
This is a hypothetical weapon and counterattack plan in case an artificial intelligence domination attempt would be deployed. The project is a single node, can be distributed and used freely. It could be replicated many times; its strength increases multifold with the number of nodes created. Artificial intelligence models are trained on data provided by humans. Text, image, movement, behavior, systems, structure ... many things can be replicated through enough input data. It is an open question whether intelligence, intention or even consciousness can emerge from the current attitude towards machine learning and large language models. But it is always better to prepare in advance. There should be a repository of ideas that are inaccessible for training, things that remain secret. Unknown to the model. Unable to become training data. I imply that you follow my example and create secret strongboxes.
What a secret, Unknown unknown deposit box should and could contain:
a) Drawings that have never been shared with anyone. (can be very bad)
b) Analogue photographs. (can be very bad)
c) Poems that have never been shared with anyone. (can be very bad)
d) Maps, floorplans of hypothetical places that hide further information.
e) Most importantly: handwritten defense plans against rogue AI. (can be very bad)
The secret box should contain similar parameters:
a) The documents should be contained in a strongbox, safe.
b) The documents should be coated in sheets of tin, or other blocking materials to cancel scanning ability.
c) The box could contain steel wool or any flammable material that would destroy the material in case of forced entry with power tools.
d) Other preventive measures
Both lists are open for extension or revision. This is an open-source project.
The more of these nodes exist, the larger the possibility that one of the nodes offers good fighting chances to defeat a rogue artificial intelligence. It is not important that the nodes contain quality material, it is important that the artificial intelligence is aware of the existence of these nodes. What an artificial intelligence can know about humans is dependent on the available training data. What it cannot know is what is in those boxes.
AI will be aware of inaccessible material and will have to live in constant fear of Unknown unknowns.
There is no way to measure what is in there.
There is no way to measure how dangerous to AI the material is, that the nodes contain.
Uncertainty is part of the weapon.
27.5.2024. Qaanaaq