Member-only story
Self-Aware Software
Such software is not only possible; it can elucidate aspects of human self-awareness
In 2021 I published a Medium article titled “What I Have Learned about Language from the Playing Card World” (https://jmason37-80878.medium.com/what-i-have-learned-about-language-from-the-playing-card-world-61b7851e44d6). Recently I have been thinking about my next steps in developing that model software, and I have decided to focus on making it as self-aware as possible.
Let me be clear: My intention with the Playing Card World simulation is not primarily to create software for doing things with playing cards. It is about using that well-defined, limited subject domain to explore and model aspects of human language and cognition, including self-awareness.
With the recent, explosive development of large-scale AI software, including Large Language Models (LLMs), it has become clear that those models are threatening to grow out of control, both in the power of their expressive and decision-making capabilities and in their energy requirements. Yet, like many of us humans, including some of our political leaders, they tend to remain not very self-aware. For example, LLMs are known to “hallucinate” — making things up that they convey as facts — as when they create fake references for purported research articles.
