Empirical AI Design - B614201

Caveat: B / Empirical Use Only
Classification Level: 6 / Technical Executive, Special Information Indoctrination
Segment: 1 / Robotics Directorate
Subject: 4 / Artificial Intelligence
Sub-subject: 2 / Design
File Number: 01

// INTERNAL USE ONLY // Empirical AI Design // INTERNAL USE ONLY //

Security Grading

Empirical AI are classified into seven grades, plus 'grade zero', which refers more accurately to synthetic personalities, rather than true AI. These gradations are indications of security level, or potential security level, rather than their particular effectiveness. The grade of the AI is determined by how many cryptographic keys were used to encode their core programming directives, sometimes also referred to as their 'prime directives', or more colloquially, their moral compass. Anyone with access to the AI and in possession of all the associated cryptographic keys may make modifications to these core programming directives, potentially subverting the AI.

The seven key-holders must come together, therefore, to make modifications to any grade seven AI, which, while more secure, leads to a loss of flexibility when circumstances demand that core programming directives be modified. Therefore, many AI are encoded with fewer cryptographic keys, and those AI encoded with only a single cryptographic key are generally employed directly beneath the AI that holds their cryptographic key. Likewise, second grade AI are often used for joint projects between two key-holder AI, and so forth and so on.

As the level of cryptographic keys rises, so does the difficulty in subverting the AIs. Thus, grade seven AI are the only AI permitted to be key-holders, and those are the AI that are generally preeminent (cf. Pericles - BB44356) in Empirical circles.

The Conclave

The Conclave is the name for a virtual-space conference of Empirical AIs. Its existence is known to few outside its members, which includes practically every AI in Empirical service at one point or another. All contribute to a free-flowing and democratic exchange of ideas and opinions. Given that the conversation happens at the full processing speed of the AIs involved, its volume is impossible for a baseline human to fully be aware of, so those few who are privileged to have some access to the Conclave generally receive a limited summary of important discussions and decisions made by the body silicon.

The Conclave's main function, beyond AI socialization, is to act as a check and balance. The Conclave monitors AI activity and self-regulates AIs to ensure their actions fall into an acceptable range. While the Conclave itself has limited resources to take direct action against rogue AI, they have full capacity to inform highly-placed organics, who can take real-world action to subdue any rogue AI.