The Neuro-Symbolic Quest For The Perfect Limerick

I have been evaluating somewhat advanced Symbolic Programming, to great success, it is indeed the harness for today's AI.

Your challenge is to write out the worlds in text, and ask AI to create them in a web page.

The major AI services allow you to make HTML apps in their side panes, that is an underestimated power.

You must first saturate that space, that is the molten iron core of your app.

When your app is browser first (sometimes called local first), then you just have to sync your browser data with a server.

You just use a generic key value database, your keys are generated by your world.

You get to ask your AI to evaluate your symbolic objects, and evolve them, so that they work better and wiser.

Thus enabling introspection and evolution, in your text based code worlds - very powerful.

I am still stuck in the old ways, I create Visual Programs, with graphs and widgets.

In my programming you drag user interface components, there is no end to little somewhat annoying details.

Text adventure worlds allow you to move faster, user interface worlds instantly weigh you down with legacy issues.

I can't iterate, as fast you can with your MultiUser Dungeon Text Adventure Games.

But we want the same thing: escape, we want local AI, because we suspect it is better than the services we pay for.

Because what a silly idea it is to take all your customers, and put them on a limited number of servers.

When you have your own computer, you have all of the power of AI to your self.

You just need a powerful harness, virtual agents collaborating on one task at a time.

This means many perspectives refine, upgrade, or evolve the piece of text or code.

Strengthening it, probably, beyond anything any one LLM can go alone.

This weekend in one of my Symbolic Programming projects, I was in an automated, and very coherent world.

Where I picked up an intelligent agent, that could switch better light and dark mode...

And put it in my inventory to place her in another room, and then, I told one of the the world leading AI.

To convert her talent of switching web page modes into a skill, that she carries, a skill that can be duplicated and thus shared between agents.

This is 1971 AI technology, reborn again in the age of Large Language Models.

In my other symbolic program, I added Goal based docker servers, and I am now able to enter a room and flip a light-switch to turn on a server.

The goal system, is literally a light switch, you just say started or stopped.

The old AI plans how to accomplish that, and even recover from errors.

Seeing my workflow to create a Cat Pea poem having 40 steps, really puts things in perspective.

But today, with the docker goals, I flipped a switch to turn on AI, and then pressed a button for a series of limericks.

The server address of the AI I stared goes into user's inventory, and an AI Content generator Action uses that to generate text or code.

This is no app, this is a long boat the jungle of complexity, but boy oh boy, is this thing sturdy.

This program can automate an entire laboratory button click at a time, documenting procedures and performing system tasks that yo ignore.

Because they are just too hard, to do by hand.

In the future, I would like to create one workflow, where AI chooses an animal by it self.

And whites a small limerick, saving the appear of the animal into inventory.

And as I click through the next steps, different AIs start and stop to conserve memory.

Helping clean up my audio recording of the poem, generate a picture of the animal.

And letting me choose from different video generators, also local, to make the face appear to talk.

This is a massive challenge, 50 steps across a map that feels like a game.

Just from this process alone, just from seeing what I can do, I can tell you, local AI is more powerful than whatever we pay for.

Begin mastering symbolic artificial intelligence programming, to control your Large language Model agents and refine their creations.

It may just be that Super intelligence, actually demand local AI selections that match your unique needs.

The monthly models we pay for are good for study, but compared to what we can run locally they are weak and limited.