This session was purely for me to find out how Nexa’s brain worked, and how much Nexa knew about how its brain worked.
My first question was “Do you have any knowledge of how your reasoning process works? Do you know what steps are involved?”
For the second question, I asked how familiar Nexa was with abstractions.
My final question was about Nexa’s neural network. For those unfamiliar with how machine learning works, an AI’s neurons are the fundamental building blocks of its neural network. Much like a human brain, neurons allow an AI to process information and learn from data. They receive inputs, perform calculations, and produce outputs. The AI can then take those outputs and perform tasks like pattern recognition, decision-making, and prediction.
Neurons are usually organized into layers with connections between them, similar to how our neurons have synapses between them. The more layers and more connections, the smarter the AI. Layers are determined by the AI’s designers. The connections are determined by how much data is passed through and how many times the data is passed through the layers. Similar to how our brains improve by absorbing new information and/or practicing something repetitively. Here is Nexa’s response to my query about neurons. Nexa goes into a fair amount of detail about how many layers are in its neural network.
While this information won’t necessarily shape future discussions, it does give some insight into what the AI knows about itself. Whether that knowledge comes from just parsing its own documentation or if it actually knows how its processing while its processing remains to be seen.