Neural Simulator for AGI research and development
I appreciate your interest in this project and hope your experience will be positive. There are many ways to participate at all levels of AI ability. As a first-timer, you can download a release, try things out, report any bugs, and make a few suggestions. At the other end of the spectrum, you can dive into the code, create modules and training sets. and more.
Brain Simulator II is an open-source, free package which simulates an array of simple "neurons" coupled with "modules" which allow any group of neurons to be backed by code.
Strategy: Brain Simulator II is an experimental platform. The point is to be able to try out different AGI algorithms and architectures with a minimum of effort to see what works well and what doesn't. It's not intended to be an application or development toolkit. There may never be a stable release, there may never be fully-debugged code...so patience and a degree of expertise are required. Development is incremental. Write a single bit of functionality, try it out, see if it makes the system better. While there is no grand plan, there is an overall strategy of a set of modules which may result in an AGI system. Accordingly, the modular system can continue to work even if other developers include modules which aren't functional yet. In fact, the project includes several modules which are not currently useful but which may prove useful in the future.
Neurons and Synapses: Neurons are tiny blocks of code interconnected by synapses. A synapse is owned by a neuron and connects to a target neuron with a weight. A neuron has a single function...it can fire. When it fires, it sends a single-value signal to every neuron in its list of synapse targets...and the message is the weight of the synapse. There are several neuron models.
Modules: A module is a rectangular array of neurons which is controlled by code. While the neurons are still functional, the code can easily perform calculations which would be tedious in the basic neuron models. Each module can have a custom dialog box and there is a custom add-in file for Visual Studio which makes creating new modules easy. If you create public properties in a module, they will automatically be non-volatile and will be saved and restored with the network.
Networks: A design of neurons, synapses, and modules is saved and restored as a unit. You can create as many different networks as you like…they are like documents which can be opened, edited, saved, branched, etc. They also save the state of any public variables in the modules so memory content is preserved from run to run along with the position and content of any dialog boxs.
Again, thanks for your interest! Charlie Simon, original creator of the Brain Simulator