By Roy F Rada, MD, PhD
Placed on LinkedIn: April 22, 2024
Keywords: Generative AI, Peace, Autonomous, Living, Multi-Agent Models
Preamble:
The previous essay (AI States and Peace: Autonomous to Living) had four sections: Automated versus Autonomous, Living Systems, States, and The Future. It provided the conceptual overview, and this essay explores implementation.
The discovery that ChatGPT talks like us precipitates a change in basic assumptions. Those who are twenty-years old cannot imagine the shock experienced by some of us who are over seventy-years old. In my computer science PhD work in the 1970s I built small, biologically-inspired systems that learned from sensory data. In the 1980s, to get research funding I had to shift to augmenting semantic nets which had become the new AI paradigm. Large neural networks, like ChatGPT, have returned us to the biologically-inspired paradigm with one shocking twist:
A computer program can leapfrog millions of years of evolution by reading all the documents written in the last few years.
The training of ChatGPT requires libraries of documents that in turn require organizations which have evolved over thousands of years. We move focus from the human to the organization. In the Modern Turing Test (www.technologyreview.com/2023/07/14/1076296/mustafa-suleyman-my-new-turing-test-would-see-if-ai-can-make-1-million/) an entity is given $100,000 and asked to generate $1 million in profit by discovering a new, marketable product, making it, and selling it – that entity is an organization not a person.
Shifting Focus of AI Research during Half a Century
An artificial organization begins with a charter and then develops a complex plan. In addition to software, it needs other components, such as sensors. Of the many types of organizations the top one is the state due to its special powers. LLMs can help the construction and support the running of organizations.
A profit-making organization needs a business plan that goes from vision, to strategy, to objectives. In addition to a plan, an organization employs operational frameworks, such as the popular Open Group Architecture Framework. That Framework assists in creating and managing resources and processes and in designing, planning, implementing, and governing enterprise systems. The Framework incorporates different perspectives, such as data flow, workflow, and money flow. Numerous subsidiary models that support organizations can be utilized, such as Role Based Access Control models that decompose the activity of the organization into roles that, in turn, incorporate functions and operate on messages among roles. Reuse is crucial in organization building, as in software engineering.
Snippet of a Role Based Access Control model
A robot
is programmable,
senses and manipulates the environment, and
consumes and acquires resources as a requisite to continued operation.
An adequately flexible organization would want control over robots. An example of an artificial autonomous organization is a fleet of cars and their charging stations where each car transports people for a price and recharges itself with the money from its passengers.
A series of prompts can direct an LLM to create a software requirements document for an artificial business. For instance, with appropriate prompting:
The LLM can examine data about businesses and determine which ones are growing and what new ventures seem likely to succeed. From there the LLM can suggest a scope for a requirements document to create an artificial business (www.researchgate.net/publication/377067114_Large_Language_Models_in_Enterprise_Modeling_Case_Study_and_Experiences).
Next the LLM can analyze the market competitors and model their vision, infrastructure and process needs, and sources of income and costs. Those models can be integrated and modified to create a new model for the new, artificial organization.
Along the way, the LLMs can critique its own results and develop refinements.
From the requirements document, the LLM can create a design, code, and test.
As matters stand, many interventions by humans might be required, such as registering a legal entity with authorities, but LLMs could be invaluable colleagues in the entire venture.
A separate issue is the incorporation of LLMs themselves in the artificial organization. Popular specifics include responding to customer queries and supporting internal documentation. More generally, a LLM can support total quality management by monitoring the performance of the organization, detecting failures, and suggesting improvements.
How might we encourage rapid prototyping of artificial organizations? Three different approaches for profit, non-profit, and state include:
For profit organizations respond to venture capital fund . We could capitalize such a fund and solicit meritorious proposals that generate artificial organizations designed to make a profit.
For the non-profit category, we target objectives in healthcare and education and seek funding from external sources to grant to proposals to create artificial, non-profit organizations that are self-sustaining.
For the state category, we seek special connections with a state that wants to develop artificial states that are satellites of it. Additionally, we get involved in existing multi-player games, such as EVE Online, to stimulate the creation of virtual states.
The preceding are practical next steps for implementation based on current understanding, but since we do not know what will work, these implementations are experiments.
We need research about on tools to evolve organizations. We have mentioned Large Language Models and multi-agent systems, but other tools will be necessary. One example of such a tool is a universal fabricator (designingreality.org) that can build any physical device, including itself.
States are the only organizations whose charter supports the use of lethal force in the attainment of its goals. These goals include collecting taxes to support police and military. Police enforce tax collection, and the military maintains the state boundaries. States try to control any technology that impacts the state’s tax collection and enforcement abilities. Thus, to create artificial organizations we must cooperate with the state.
Two customers with money buy corn from one store, and the other store fails. Dollars from citizens to government.
Paradigm shifts spark revolutions and dreams for salvation. Peace among states is interrupted by wars because evolution drives the intrinsic competitiveness of states. Artificial states could enforce harmony among states on earth and grow resources through new opportunities, such as the conquest of interplanetary space.
