The best chat on AI Agents until here.. is here...
The trick is to use the LLMs only as a translation layer to trigger rules, that is in AI terms mixing a Large Language Model with a Rule Based System and thereby a form of a Neural Network.
Neural networks are machine learning models inspired by the human brain, using interconnected layers of artificial neurons to process data, recognize complex patterns, and make decisions without explicit programming. As the foundation of deep learning, these models train by adjusting connection weights, enabling applications in computer vision, speech recognition, and natural language processing.
Neurons (Nodes): The basic units that receive input, process it, and send output to other neurons. (LLM)
Layers: Neural networks consist of an input layer, one or more "hidden" layers, and an output layer. (OS and App rules)
Weights & Bias: Connections between neurons have weights (strength) and thresholds (bias) that are adjusted during training to reduce errors. (log database)
Training: The process of feeding data through the network, calculating error, and using algorithms like backpropagation to optimize weights. (log analyzing by eg a reasoning LLM)
I call that AI Ghosts, because most of what makes them you wont ever see.... here....
#cyberpunkcoltoure
