3 ms·
I’m not sure your rebuttal is sufficient. Weights and biases in the form of a neural network aren’t static either. Deep Reinforcement Learning algorithms exhibi
by blovescoffee 5y ago
I’m not sure your rebuttal is sufficient. Weights and biases in the form of a neural network aren’t static either. Deep Reinforcement Learning algorithms exhibit learning and growing.
Besides who says complexity is necessary or that ANNs are not complex enough?
- tasty_freeze 5y agoThe idea being discussed is that a person could live on if the connectivity and weights of the neurons in their brain was captured. "You’re just a set of weights and biases right?" I pointed out that a brain is dynamic; it depends on the biology of that person, and there are many other control loop mechanisms which are not captured by a network of weights. We aren't "just a set of weights and biases". As practiced today, neural networks are trained, and then the trained network is statically deployed in recognition tasks or whatever. They aren't continuously evolving. Yes, one can come up with algorithms that retrain and reshape a network continuously, but (and here is the key point) it is very likely to quickly diverge from the changes that the original, organic brain would have made.