Category: Agents
Updated on January 3, 2021
The OpenSense Domain
Communication is essential in allowing individuals to cooperate in group activity, especially if the individuals differ in their roles or characteristics. Having open access to information greatly boosts the productivity of a group – in fact, this is the motivating principle behind the invention of patents, the internet, and open source software. Similarly, members of a team are expected to openly share their thoughts and tendencies so that the team can make up for each other’s weaknesses. In the OpenSense domain, we discuss what happens when important facets of generic thought are openly expressed and easily sensed by others in the environment. We will talk about how generics in this domain tend to form friendship groups, share information about themselves, and engage in play activities designed to uncover highly varied aspects of each other’s personalities. The senses in this domain are analogous to human emotions and the involuntary facial expressions / body language used to express such emotions1, but in this post I will mostly focus on the OpenSense dynamics in its pure form and only use human behaviors as illustrative examples.
Posted on October 14, 2020
The Rivalry Domain
Previously, I’ve been talking in broad terms about very general concepts in proficiology. This will be the first time I narrow my focus into a specific domain. A domain is a restricted setting with only a small number of relevant events / senses / agents – basically a toy problem, or a simplified model environment. The hope is that we can use important ideas from these domains as fundamental building blocks that will help us analyze more complex setups. In other words, we should be able to build more interesting / realistic generics by mixing & matching simple components from multiple domains, as if we were building a complex molecule atom-by-atom.
In this post we will talk about the Rivalry domain, which focuses on events where one generic benefits at another generic’s loss. I will supplement my explanation with formal notation loosely based on functional programming languages. I am not requiring (or expecting) readers to have a background in computer science – it’s just that this kind of notation is very useful for describing nontrivial generic behaviors through the composition of simpler constructs. In any case, I will be explaining this functional notation as I go.
Updated on October 12, 2020
Agents – The Source of Motivation and Action
The last two posts2 were focused on the lens. They discussed its various modes of operation, the dichotomy between the outward-facing “causal estimate” and the inward-facing “evaluation”, and the problems of foreign context and overgeneralization.
But it’s important to note that the lens is, at heart, a statistical machine. It is only concerned with how accurately it can perform its tasks (casual prediction, event interpretation etc.), and does not intentionally distort itself to satisfy any agenda. In other words, the lens lacks agency and a hypothetical “lens-only” generic is only good for making uninterested predictions or classifications3.
To breathe more life into the generic mind I introduced the agent, a cognitive process with an inherent goal to cause the production of certain evaluations. The concept behind its operation is as follows:
- Use the lens to perform value prediction. In other words, given some known prior information and the desired evaluations, figure out what must be in the posterior information for the lens to produce such an evaluation.
- Try to engineer the actions and circumstances of the generic to increase the odds that the right events happen. If this is done successfully then the lens will produce the evaluations that the agent desires.
