Your Own Artificial Intelligence

Today’s optimized and reduced models, are two gigabytes of ram, they lack detail, faces are distorted in photos, for one.

But they do reason, so is you place them in a conventional AI harness, that ask the right questions, they will work.

The very big AIs get their knowledge from their brains, but you can make a local AI very powerful too.

And if you add the multiplier of specializing it at something specific, you can do great things with it.

A 2,000 computer at the local supermarket comes with 12GB video ram, and a 196GB video ram remote server can be rented at $6 per hour.

And then the world’s smartest AI costs $20 per month, and that will do all the programming you need.

To be specific, you need to program the conventional AI, that marked a breakthrough in car problem diagnostics and computer games.

It is very simple, the problem is has however, is that it need a lot of configuration, something today’s AI is perfect at doing.

You can really make expert systems and goal solvers shine, by building up their knowledge with today’s large language models.

Shine at what? has two interesting answers: Computer Games, and Computer Programming, specifically Code Generation.

I’ll let you dream up computer games, but let me tell you about programming.

I learn web design ideas from a wise CSS framework called Bootstrap, I always try to make my own bootstrap, and always go back because it is easier.

But if you had time, to build up a knowledge base that generates something like bootstrap.

Classes, user interface components, web component components, and an clean XML based language far cleaner than the world leading React.

I hope you will be starteled to realize that conventional AI is, better, for precious code generation.

Like I mentioned above, the problem with the old shool AI, or as we should call it GOFAI Good Old Fashioned Artificial Intelligence is typing.

A lot of time, experts and mechanics had to collaborate on car diagnostics, today, it would take minutes, probably 5, definitely less than 15.

You have that power to build GOFAI in your smart AI, it has to specialize at something, the more specific the better.

And don’t worry about GOFAI rules being determinate, just introduce a probability function, temperature.

You can also create a hybrid, a harness like AI where rules are instrucions.

Not strict GOFAI data structures with provides/requires, that get matched up and self assemble for a solution.

But instructions for a tiny local AI. who can probably handle them shockingly well on a slow computer.

The result is local AI that runs on your computer, perhaps in a webpage even, which also means it would run locally for your customers.

You can look at electronjs showcase, find an open source applicaion, with a solid foundation, tear out their code wit AI and make a base for yourself.

Where you can house you code generator or generators, now you can sell an application that people would want, if that is what you want.

This is a starting point for your dreams, for example, you could create a Gofaistudio or Gofaishop.

Where people can plug in big AI to make their own conventional AI systems, for their own specific purposes.

Large language models are too slow, for us to think about dynamic user interfaces created live as needed.

But Gofai could not be more efficient, at procedurally generated user interfaces.

You could try building applications that don’t ship with a UI, that assemble user interfaces as needed.

Big AI is very interesting, but no one will ever take away you local one. and the local one can be even better at fundamentally precise tasks.

There is no programming knowledge required here, ask your AI to show you GOFAI examples in declarative XML, and learn form there.