Are We Neurons in Networks? 6 April 2018 Thoughts There is an interesting assumption, brain in a vat, which supposes what we see or what we feel are exactly programmed by certain computer, and we do not realize it. And the movie The Matrix also based its background on a similar situation. Today, my question is, can everyone be a neuron in networks, and what kind of networks? I grow up since I was born, and I interact with others or the environment to learn something new. It seems that my body is the concrete network architecture while my soul is the algorithm to perform. Compared with the most Deep Learning algorithms today, the differences are, my body is updated through metabolism, it may become stronger or weaker, according to the different stages of life. This process is programmed into the body itself, and not changed by my soul. But maybe, my soul always acquire new information from outside, and update the parameters in my body by producing new cells and discarding the old ones, and information is transmitted between old and new cells. And this kind of procedure is not exposed to my soul, because it is a private function and encapsulated. It seems that our bodies are more like Recurrent Neural Networks, especially LSTM, with the soul trained with strong transfer learning. If we regard our bodies, many network architectures, as many neurons in some big networks, what the reasonable explanations are, corresponding to the neural network models at present? We see each others and some same parameters in the world (But here, I am not sure if you are my only imagination or real, but let us make it easier). So maybe we are in a network similar to Convolutional Neural Network, and we are sharing the same filters to understand the inputs from outside, and no wonder we will hold different thoughts on the same things, since we are initialized with different values (reference DNA). And even we are born with the same DNA, we can still be different because of mini-batch stochastic gradient descent. We can compete with and learn from each others. So are we in a network similar to Generative Adversarial Net? The difference is, everyone can be a generative neuron, facing thousands of discriminative neurons, and vice versa. Since we form some big networks, what are the outputs of these networks? Take the school system for an example (it may be one part of a big network), the neurons in this part are students, teachers, and other staffs. Student neurons are not trained well at first, then they receive information from teacher neurons and other staff neurons, absorb it, output performance, and adjust performance by interacting with others in a short time. It is not an end-to-end fashion, but a multi-stage learning process. Finally, the student neurons may have the ability to output good results from input data (e.g., input a math question, the student may output the right answer). But what are the outputs for the society network? For country network, it may be comprehensive national strength, sometimes trained with jointly learning or GANs. For Earth network, it may be a better human living environment, by uniting all the neurons in the world with different wights. And throughout human history, human reproduction and historical replacement are only parameter updating of this network. So, who is the creator of this network?