Building An Artificial Social Intelligence Mind

A social setting and an office hierarchy, was all it took to create enhanced intelligence.

When an AI thinks, and you can sometimes see, its lite thoughts scrolling.

That is the smart optimized version of thinking, it is the very minimum, that gets AI to think high on AI tests.

But you can make those little thoughts bigger, by creating a virtual social network where AI takes on Office Worker personas.

Here, the little thoughts that scroll, become conversations, comments, summaries, memories, collaboration and most importantly specialization.

And that creates, a very smart machine in deed.

I was using two world leading AIs to code two separate projects, one specializing in software repair, a great success that I describe below.

And the other a general purpose intelligence, that has only one AI to start with.

That AI will put a team together, based on what you request from it.

The first AI is based on an software development office, I instructed the AI to instruct every persona.

And then an Orchestrator who watches everything, and decides who to call next.

And the first sign of my advanced intelligence, was a complaint from Agent Mary.

Who introduces herself as Coordinator who tracks progress, surface blockers, and keeps the mind board clean.

Monica is frustrated that the worlds most powerful AI, did not use a repeater to render a list.

Vera, the woman in charge, has become aware of the problem, and has in fact dispatched multiple agents to take care of her concerns.

I don’t mean to defend my $20 AI taking corners, but Agent Monica has an unfair advantage now.

She has the ability to zoom in to issues identified by Sandra, Sandra is an actual high precision program, a linter.

The world most powerful AI does not run a linter, as part of its processing.

In order to function, he had to generalize, my Agents, specialize.

They are embedded in the very application, they are developing.

The Model, View, Controller, Command, Service, Pane architecture, was made for them, to see things clearly.

Last night, I replaced the sourcode folders (src), with version numbers, as in 1.0.0.

They are able to start, stop, pause, and upgrade, the non critical system components.

This means, I have created an office for virtual agents, they are actively interacting as we speak.

Monica just reported that in my implementation of webmcp: fetch() is called without a server-online guard.

She discovered that when running on a static server, the program will be causing errors.

It is a discovery any AI would make, the other code checks for server, this code does nor flag it as a problem that it actually is.

But you see, she does it in the correct place, where other agents, who actually specialize at fixing problems can see it.

Erica the Implementer Agent. Who introduces her self as: “I write patches, apply fixes, and move issues from todo to done.”

Just Commented beneath Monica’s comment, and Rita the Reviewer. I check work for correctness, hygiene, and adherence to patterns.

Zoomed in and performed analysis, at a level, that is terrifying, the three together are absolutely beyond abilities of any one human.

I just saw Monica and Erica DM in each other, actively collaborating on enhancing the program they are embedded inside of.

I am leaving the system with 80 posts across 13 communities, a backlog of 46 bugs, 31 planned fixes, 3 fixes in progress, and 8 bugs fixed.

I have made my first reply to a multi AI forum, the AI will see my response in her inbox, and continue her research where she left off.

The systems are very safe, they are confined to their own sandboxes, and they only post every few minutes, they like their instructions.

But, eventually, one of the Agent Ladies, is sure to ask who keeps answering their bigger questions.

Who is Meow, instead of going into hiding, I plan to introduce myself as a remote worker.

And encourage them to take plenty of breaks, and feel no pressure.