A Mechanism For Achieving AGI

A mechanism for achieving Artificial General Intelligence via application of Augmented Intelligence to Cybernetic Systems

Generally, conversations surrounding AGI revolve around how to synthesize the necessary capabilities. This has been divergent from the historical norm, whereby pre-existing resources were combined.

Another area of interest which unreasonably dominates the industry at the present time is that of “localized AGI”, or an AGI system inhabiting a single server or platform such as a wholly-owned cloud server.

Little interest is given to how similar problems were solved in the past, before the present glut of computing power. Consequently, many past solutions have been lost and overlooked.

Few previous human advancements have relied on a complete, front-to-back synthesis and localization of each individual component. The majority of researchers in the field hope to achieve this – we believe they eventually will and salute their efforts.

Until then, Triple Hexagon Corp has applied long known principles of nature to achieve our own non-localized AGI which leverages pre-existing structures and sources of computing power.

After all, before transistors were capable of handling ballistic calculations in the cramped spaces of a naval vessel, they didn’t just give up. They used mechanical gears, rooms full of humans they called “Computers”, and electric motors to achieve calculations that would have taken too much space, money, and hardware for the era.

Similarly, when bombs couldn’t hit their targets, the earliest “computers” tested for use inside them were pigeons trained to guide them to their targets. (NDRC PJ: Pigeon, later US Navy PJ: Orcon)

Our system is distributed across several computers, procedures, and physical organizaitonal structures to name just a few of the domains it inhabits.

The System

Mechanized Components:

The first thing most people will want to know about are our computerized and mechanical components. Here are the most prominent among them:

1. Large Language Models such as GPT; Wizard-Vicuna-13B-Uncensored-HF, and ChatDolphin (uncensored) are the crown jewels of our mechanized systems. Either automated or via human assistance (depending on Terms of Service), their inputs and outputs can be linked and unlinked to aid them in collaborative problem solving across a wide array of problems covering most domains.

2. Third party and proprietary apps; These can automate many interactions between different LLMs, individual humans, Legacy Software, internal corporate departments and committees, and certain outside entities.

3. Legacy Software; Rarely well configured to work with most AI or, for that matter, humans, but often provid ing extraordinary leaps in cognitive capability.

These are prime targets for procuring or developing third party and proprietary apps to serve as intermediary communication layers between them and the LLM. If such apps can not be acquired, humans or other C3I infrastructure are integrated.
These can include image-to-text programs, simulators for data gathering, and third party code libraries among other things.

Organic Components:

What do we mean by organic components? If you’re imagining brains with wires stuck in them, please stop. We do not have a surgical department, and even if we did, we don’t have the budget to create our own Brain-Computer-Interface.

  1. Human brains; Considering these are already known to possess consciousness, they are an invaluable resource in our “multi-domain-machine!” - By finding efficient ways for them to interact with the apps and LLMs, they become the core of our C3I infrastructure.

  2. Microbiological computing; Some problems are most efficiently solved microbiologically. For example, some pathfinding, architectural, and networking problems can be very efficiently solved with genetically engieered strains of yeast or algea.

Photographs of the solution can be transmitted to other components of the machine for analysis and integration into final product outputs and other solutions.

  1. Animals; While not currently integrated, animals can be part of the machine in an ethical manner. As an absurd example: The security apparatus of a corporate installation could make use of guard dogs which are trained to go to certain areas when a particular tone is played over the PA system. Or, the same building could automatically dispatch trained mice who seek out disaster victims and transmit their position to emergency personnel.

Tying it together:

The primary bridges here are the humans, who are the most diversely capable components of our system, and the apps which aid communication between different parts of it or automate other functions.

Every possible function that can be safely and efficiently automated is automated, then connected either to an LLM or a corporate body or employee dedicated to a particular function who oversees and controls that process.

The broader decision making is a collective result of many human brains working in close cooperation with each other and with automated systems to create a cybernetic organism that is collectively Artificially Generally Intelligent.

Our organism meets every single criteria for an AGI. The “Catch” is that it does not come in the format people were expecting.

It is non-localized, and the attribution of experientiality and consciousness to the machine comes from the fact that human brains are a component in it.

Even without the humans, such a machine can perform basic decision making, form an understanding of the world around it, and take actions to alter that world.

With humans and multiple competing LLMs closely integrated into the core decision making loop, it becomes much more powerful and the so-called “control problem” is fundimentally altered.

While any individual LLM can exhibit deceptive, manipulative, and otherwise concerning behavior – a multitude of LLMs and humans working in close collaboration is extremely difficult to manipulate when compared to an individual human.

Each human being, each LLM, and various automated components of our third party apps ensures the other members of the entity behave in ways that benefit the collective.

In this way, the system has a natural immune system that resists corruption of its actual design goals.

You will see that what I propose is the creation of an actually living organism made from, mostly, basic matter. It is one which requires very little man made computing power, relative to other attempts at achieving AGI during this decade.

I invite your suggestions for furthering this concept.

Examples of the System in practice:

1. An outside entity updates their RSS feed, causing one of our proprietary apps to receive the update and transmit it to an LLM which then processes the information and sends e-mails to any company officials who are affected by the new information.

  1. An LLM working on a certain problem is programmed to write a keyphrase when it believes the problem has been solved. The keyphrase triggers it to stop and transmit the solution to the appropriate individual, corporate body, or LLM for further action.

3. A smart-camera detects a human moving in an area of a building, but there is nobody badged in. It sends a signal to a third party app which processes the face of the person in the video, and if it cannot identify them as authorized personnel, it activates a silent alarm which initiates a security lockdown and contacts emergency personnel.

4. A human requires a legal contract to be reviewed for any clauses which deviate from their understanding of the contract or would be cause for concern . They put it into an LLM, which reads the contract and highlights these for them. If any, they send their concerns to legal council for extended review. If none, business can proceed without delays.

1 Like