Future of Life: Preparing for Wisdom in the Nhà | Part 11 | (#23)
Series: Dear Chris
Dear Chris,
We regularly choose to perceive the world in a way that will help us to better navigate reality. We must do it again.
Once, some believed we could know more about “here.” Create information about geographic space that was more real, more useful. Humanity wrapped Earth with the idea of a global coordinate system.
I believe we can know more about “us.” Create information about Human Values that is more real, more useful. Humanity can wrap our world with the idea of human values.
With this series of essays (“Future of Life” Parts 1-13), I present a perception of reality that formalizes the meaning and usefulness of human values in ways to help resolve the challenges of AI alignment and equitable power distribution.
I presented three theoretical foundations necessary to create a human values-oriented perception of reality fit for this world and the one we want.
With a new perception of Human Values and how we relate to reality, each of us and our species can begin to notice our world anew and better describe, assess, adapt relationships between actions, values, energy (potential power) where we must (e.g. companies, nations, powerful individuals).
I’ll bring together key points from this work to complete an application to the Future of Life Institute’s grant program “How to mitigate AI-driven power concentration.”1
Chris, why am I going to all this trouble to submit an application that is unlikely to be successful?
I need to know if I should scrap this work or put much more energy into it.
This is a little bit all-consuming. I don’t sleep too well. I want encouragement, discouragement, direction. I need to move on to the next stage. I’ve been thinking and working on these topics for 15 years and always.
If I am misunderstanding the situation. If someone has done this before or knows why I’m wasting my time. If my work is irrelevant, unimportant, duplicative, misguided, I should stop.
If a theory of human values is important, tell me to work 10x wiser.
What will I propose to the Future of Life Institute?
I’ll ask them to support the design of a bunch of big projects, a portfolio of applied research, that begin to systematically generate learning for and draw attention to a Theory of Human Values, a Values Layer, Values Signatures and related theory, methods, tools, ideas.234
The list of projects:
Inventory of values.5 6 List all values. (e.g. beauty, fun, friendship).7
Inventory of representations of values. 8 List all representations of values. (e.g. sunset, game of tag, postcards, people).
Inventory of patterns.9 List all perceived patterns of organizing energy.
Bounding. Ways to bound values, representations, patterns, agents. 10
Matching. Ways to connect values, representations, patterns, agents.1112
Thermodynamic Accounting. Mapping values-energy relationships.13
Energy Agents.
Who or what controls energy, how much, how, when.
What values are assigned to intentions, actions, outcomes, people, groups. 17
Integration. Mixing the values layer with other layers of information for specified uses and outcomes, e.g., geographic, economic, health, demographic.
Privacy, Safety, Security. Protecting Humans, Identity, Humanity, Values…the most powerful forces on Earth. 18
Economics. Alternative economic models.19
If you are thinking “Oh. My. That is a lot.” You would be correct.
This portfolio of work and the underlying theory is designed to enable communications, Alignment, Agency in the Nhà. That’s a lot.
10 years ago, I’d have guessed this research portfolio would take 100 university researchers 15 years to complete. And they would fail. Now, I’m wondering if 5 or 10 full-time, supported by solid AIs, institutions, advisory groups might knock it out in 5 years. And not fail.
(Not that I am any good at guessing this kind of thing)
We need a portfolio of work, a direction, initial conditions that enable Humans and AIs to communicate about human values. That enables AIs to communicate about values and humans to be able to perceive and assess their communications and related actions. We need to figure out ways for Humans to speak about our values as individuals and as a species if we are to have a hope that we can communicate our values with other species (i.e. AIs) much less give that species power over our values, our identity.
We need to be able to assess and achieve moral coherence. We need to perceive where AI power is aggregating, how much, with whom, how, why. We need to perceive the relationships between increasing Human power on Earth and the intentions and agency of Humans and Humanity. We need the ability to make wise judgments about where power should go to create and sustain equity and balance.
However long and however many people work on it, it will take a lot to transform the perception used to create modernity into a perception fit for creating a future we want.
This work represents a species’ choice to take the first tiny steps of an epic journey toward a more purposeful and systematic evolution. One for the well-being of all.
Eratosthenes and others were inclined to believe they and all of us could know a lot more about “here” and “there”.
Nothing about imagining and building a global coordinate system, a global layer of geographic information, was linear, step-by-step, well-planned, highly predictable.
Humanity wrapped its imaginations around the physical world. Captains used imaginary lines and coordinates to traverse the seas. They didn’t suddenly safely sail the most efficient routes. Eratosthenes did not immediately get his beautiful Sicilian wine. Postcards were lost. They still are.
From unconscious, informal, automatic, natural, instinctual awareness of “here.” To explicit intentional attention to questioning, designing, understanding of “here.” Humanity traversed a 500,000-year circuitous path to design and adopt a perception of geographic reality.
Today our entire world is constructed on top of and rooted to a perception of reality where everything is somewhere and everyone can know where all the time. Few of us care about how or why this happened. Even so, this perception of reality has become a feature of our consciousness. Our identity.
We don’t know how powerful AIs may become, who will control them, how much autonomy they will have. We don’t know how to assess what, how, who’s values they will serve. We do know what impact advanced AIs will have on everything or anything.
We should assume the impact will be massive all around. We should assume the rate of change, the evolutionary cycles of our species, are no longer geologic but now circadian.
Most of us would prefer that the power AI adds to Humanity is distributed equitably rather than concentrated with a few nations, companies, individuals that control the rest of us.
Humans have, forever, been aware of human values. All of us live according to values for our entire lives. We have been allocating energy to values forever. Values are the most powerful forces on Earth and eventually in the galaxy. 20
Human values presage a future of infinite abundance. But we are infinitely ignorant of these forces - the relationships between values, energy, time. That can’t be good.
We haven’t had a good reason to get serious about understanding these forces until now.
And now, our well-being, our existence, depends on it.
https://futureoflife.org/grant-program/mitigate-ai-driven-power-concentration/
Many ways - models, methods, tools - may be appropriate to begin the investigations.
All below items are assumed, at their limit, ultimately to be most beneficial, to be global and comprehensive. Testing anything from the portfolio below requires wise selection of starting points, sample geographies or demographics or individuals optimized for learning about the given method or tool. e.g. starting with samples where values have been studied most, with most easily available credible relevant information.
The following models, methods, tools related to a new perception of values, a values layer, values signature must be assessed with many and diverse criteria: Public good, privacy, security, accessible, useful, credible, secure, transparent, adaptive, sustainability, testable, verifiable, validated, useful, adaptive, credible, transparent, for well-being.
I use “inventory” to keep it simple but something like “dynamical relational architecture” is a better fit.
Basic but globally comprehensive foundation architecture of graphs (eg ecological-social graphs) using organizational, legal, and literature/media to document, triangulate, map values using existing information. Variables related to values might include: definitions, descriptions, diversity of perspectives, geographic, demographic, individual-social, dynamics, evolution, origins (short/long term), related values, stability, power, frequency, many.
In the “detailed project description” I will lay out some basic approaches to designing and creating this initial architecture of values.
representations may be structures, processes, ideas, knowledge, emotions, objects, living, non, just about anything that humans would associate with one or more values.
matching patterns to representations. patterns of organizing energy (in any form): matter, heat, structures, processes, transformation/allocation of. “organizing” indicates evolving, flows, dynamical, open, complex.
Methods and tools for establishing boundaries, what’s in and out, of a given representation, pattern, value. E.g. temporal (series, past/present), spatial, social (individual, group).
Matching distinct patterns to representations, representations to values.
Existing methods and data will be helpful moving from macro values layer information to more distinct, small, specific values systems, eg for a person or group. For example, World Values Survey and this research. https://journals.sagepub.com/doi/10.1002/per.2170
How much energy moving into and out of values. Speed. Frequency. Assessing values signatures and their evolution may require information like energy allocated per unit time to a given value or value system. Is btc a purer measure of this (energy to a given values or values over time)?
Methods and tools for tracking flows of energy into out of, between systems of interest, e.g. representations, patterns.
Energy per values per unit time. Changes and change over time depicting evolution of given systems. Assessing energetic characteristics patterns, of dissipative structures. This could be similar enough to combine with “flows”
Assessing relationships between any of the above. patterns/representations within a bounded system of patterns/representations. Meaning of all relationships in bounded system, between systems.
All of this is a way of breaking apart what is whole, a way of analyzing and seeing reality, the flows of energy. Not the right way but I’m only intending to start a conversation about how we might choose to perceive, not say how we should or do perceive.
This project can only be pursued if privacy and security are significant values. Without these values guiding initiatives from the beginning, pursuing this vision will likely exacerbate the situations it aims to resolve. Technologies exist that will be helpful. Methods and tools may involve zero-knowledge proofs and blockchain. [https://en.wikipedia.org/wiki/Zero-knowledge_proof].
If a theory of values, a values layer and values signatures are possible and privacy and security of values and identity also, then transformative economic models and incentives can be designed to reverse and reshape net flows of energy between people and corporations, humans and AI, nature and world we’ve built.
This is a perception of reality that we may choose to adopt for the future of our species.