Encyclopedia Autonomica March 2024 digest

Dear Indiehackers,

This month my Substack Encyclopedia Autonomica (E21A) focused on AI Agent concepts (Object permanence, State-Operate-And-Result, and Production Rule Systems) that are important but likely not well known throughout the community as these are usually hidden behind abstraction layers and paper references. We further provide an exploration into important Financial Datasets and a coding exercise for tracking Starlink satellites in space.

Let's dive in.
The Importance of Object Permanence
E21A started with a deep dive into object permanence. Here we provided an explanation of what object permanence is, why it matters for reasoning and planning in agent technology, and how this concept can help us train better agents to operate autonomously in uncontrolled environments like factories and in-space assembly.
Read > https://jdsemrau.substack.com/p/the-importance-of-object-permanence

Since reasoning and planning are at the heart of the research we are doing, E21A provides a deep-dive on the SOAR (State, Operator, And, Result) architecture.

An introduction to SOAR
The SOAR architecture operates on the Problem Space Hypothesis which is a psychological theory that suggests that problem-solving involves the creation and manipulation of mental representations (some call it a world model)
Read > https://jdsemrau.substack.com/p/a-short-introduction-to-soar

Production Systems In Autonomous Agents
Next, E21A explored Production Rule Systems and how they matter in perceive, plan, act architectures for our agents. In general, production rule systems are logic system that have two essential components:

• A sensory precondition (often expressed as an “IF” statement).
• An action (commonly denoted as a “THEN”).
  If the precondition of the production is met, i.e., it matches the current state of the world, the production is triggered.
  Read > https://jdsemrau.substack.com/p/production-systems-in-autonomous

Then we further expanded on our work in the domain of Financial Services.

An Exploration into Financial Datasets
As data is a key item on our roadmap to build an AI agent that can operate autonomously and support (hopefully) profitable trading, E21A provided an overview and exploration into several financial datasets that are relevant for training agents that are capable of sentiment analysis, named entity recognition, summarization, and quantitative reasoning. #fintech
Read > https://jdsemrau.substack.com/p/financial-datasets-a-comparative

If you are looking for a job, then you might want to keep an eye out each month for a new edition of our jobs and events post.

Code Clinic | Tracking Starlink Constellations in Space
Last but clearly not least, our coding exercise for this month was focused on propagating satellites in space where we explored a lightweight solution to track #starlink satellites during space-flight. This exercise used two popular javascript library and is a nice exercise to understand trajectories, constellations, and basic flight dynamics. #newspace #spacetech SpaceX
Read > https://jdsemrau.substack.com/p/code-clinic-tracking-starlink-constellations

We hope that our work is of interest to you.

If you have any comments, projects, or ideas, please reach out.
Otherwise, like, share, subscribe.