Integrating STEM and Ethics for AI Alignment / Goal Alignment

While some developments in AI may inspire some people to rethink their assumptions in such areas as the nature of language and cognition (such as emergent properties that were predicted to be impossible to emerge), a number of other topics are likely more deeply buried in the fascinating disjunction between the early-stage and slow-rate of the evolution of STEM related concepts (among H.sapiens-humans), and the rapidly accelerating rate of available technologies including AI.

In an area that variously gets called the AI Alignment Problem or the Goals Alignment Question (or any of many similar permutations), it is often a strangely unexamined assumption that there simply cannot be any intersection between STEM and ethics. This is perhaps an ironic cousin to parallel statements that have been gradually overturned as part of the Enlightenment project, most notably: there simply cannot be any intersection between STEM and…the workings of intelligence (that intelligence must somehow have anti-STEM origins). This cookie-cutter claim has been likely applied to nearly everything that anyone has tried to study since the dark ages.

Whatever the cause, there are many areas of concepts about STEM that need developing. One of these is the set of unquestioned dogmas that nature, science, and STEM have no possible connection to value, function, and meaning. Not only is there ample ground for machines and H.sapiens-humans to share ethics and values, but also to dovetail that with policy and collaboration.

In the absence of a better introduction, the proposed approach and solution is outlined here:

https://github.com/lineality/object_relationship_spaces_ai_ml

Does this approach make every possible concern evaporate or sublimate? Surely not. But there is a lot of low hanging fruit, and there is a lot of work we can get started on right now.

About The Series

This mini-article is part of a series to support clear discussions about Artificial Intelligence (AI-ML). A more in-depth discussion and framework proposal is available in this github repo:

https://github.com/lineality/object_relationship_spaces_ai_ml