Post #3517168

I've been training a custom model for years. I've used an updating set of training data, different training modalities, varying applications of backpropagation and reinforcement approaches to correcting errors as discovered. I've used this model to create and complete many successful software projects. Design some hardware projects, even plan driving routes, and strategies for boardgames and video games. This model has been an instrumental component to the majority of my income. The early model learning was slow and expensive, but it paid off. Even though the model is fairly competent today, I find that further training trends to prove useful without degrading it's capabilities. In fact, it seems and to use the new skills and knowledge to combine with past training to synthesize novel solutions to n current age past projects. I do allow the model periods of "dream" time. Seemingly random prompts that may be pulled from past exchanges presented with different contextual components.