Support Independent AI Research and Help Make History!

Hi!

Last summer, someone asked me a set of very challenging questions about the ways childhood experiences affect our experiences of events in adulthood.

Those questions catalyzed an AI research project that had been on the tip of my mind for a few years.

In fall 2021, after mulling those questions over for months, an answer and an experiment hit me like a lightning bolt, while I was out walking one evening: I recognized how attachment styles could be interpreted as quantifiable learning patterns, which emerge as optimal under different environmental conditions.

That insight began my journey as an indie AI researcher. Once I recognized how to combine attachment theory an Q-learning in a single coherent framework, I couldn't do anything else. I'd seen a way to describe how experiential factors shape our interaction styles -- and what it takes to shape our responses and our environments, in turn, to foster greater happiness and health.

(In other words -- no attachment style is wrong, and nobody is broken, and mathematics can guide us in making choices that improve the quality of our lives: immediately and intergenerationally.)

Before the year was out, I'd built out code that expanded an existing standard library for AI research in order to test three different things systematically:

1. Could the evidence-based classifications of attachment styles -- categories that characterize our patterns of interaction in intimate relationships -- be described mathematically as consistent trends in the hyperparameters used by Q-learners who perform the best under different sets of environmental setups? If simulated learners showed patterns that map to the interactions patterns people show -- like our inclination to explore, our responses to change, and the value we place on long-term rewards -- that alone would be a meaningful contribution to human understanding.
2. Can we determine quantitatively which environmental conditions affect different kinds of optimizations?
3. If attachment styles do emerge as optimal learning strategies -- is it possible to learn new attachment patterns long after our earliest formative years, by choicefully shaping our environments and our responses in ways that promote stability and health?

I was pretty sure I had a way to describe a general solution to the question that had moved me so powerfully -- and with it, the possibility for showing how people could better understand one another's backgrounds, to foster one another's strengths, and build a more compassionate world. It was a mathematical foundation for becoming a better person, that I had to make real, and share.

In the last two weeks, my efforts began producing visible, compelling results. My hypothesis seems to be spot-on. Part 1 is nearing completion, and the other parts are underway.

I'd like your support to sustain this work, and take it to the next level.

This exploration of social complexity science crossed with AI is "basic research": it lays the fundamental groundwork for future developments in psychology, social complexity science, and policy. It could lead to individual decisions that create happier lives -- and support decisions on a society level that create better schools, foster care systems, neonatal care units, and even healthier corporate cultures and correctional systems.

It's a topological map of compassion and wisdom, with the potential to improve the quality of life for countless people.

These are my asks:

$4500 to cover AWS compute services.

(So far, this project has accrued $2960 in compute costs. Based on that experience, I'm confident I can complete it for another $1500.)

$3200 to cover living expenses while I finish. I work part-time as a science & stats tutor, but making ends meet while doing research instead of DS client work has been hard.)

That's it. If you can help me keep the internet flowing for a few more months, I'll deliver a mathematical theory of attachment landscapes.

Personally -- this project represents a synthesis of my skill gained from 8 years of work in ML, my training as a physicist, my practice as an ethicist, and my commitment to lifelong learning. It's also an opportunity for me to establish myself as an independent researcher after a couple of very difficult work experiences involving the violation of my civil rights -- an all-too-common experience for the tiny number of AI researchers with backgrounds like mine, who have come as far as I have.

This work is not merely intersectional. It's a nexus of several frontiers in human development.

With your support, I'll demonstrate resilience, dedication, and skill, and turn my potential into reality. I will manifest a scientifically-sound opus of love: a contribution to the wellbeing of my loved ones, my species, and to all beings on this planet who care for one another.

-- Khayah Brookes