Now imagine that instead of modelling a physical neuron, we create an approximation of one using a neural network that has been trained to approximate a neuron. This can be done over time for every single type of neuron, this is a slow method, but I cannot imagine it being slower than physically modelling a neuron. Physical modelling is difficult enough that even a naive implementation of matrix multiplication, as would happen in this "neural" network in O(n^3) would probably outpace it significantly. I think that this model might resemble a multi-in-multi-out version of a perceptron, with some sort of approximating function to determine synaptic reconnections and topological changes. The overall network layout may very well approach something like a Convolutional-Neural-Network, or CNN. CNNs are the sort of network used by the google image searching functions. They are very interesting and have a lot of improvements to be made on them before I'd trust them to simulating me. But I trust that ensemble methods built to hybridise this network type with additional machine learning methods such as decision trees or any other sort of model that you'd like.
Alternatively we can have a genetic algorithm if you'd prefer, that would model more processes in the brain and eventually lead up to a high fidelity simulation using the minimum(this is what evolutionary algorithms tend towards) amount of complexity to achieve a level of realism. This is a tricky method to write because sometimes genetic algorithms end up finding optimal solutions for only their fitness function, so these minds may have various quirks that may prohibit them from being... normal I guess.
Fuck all knows if these models can handle learning in a coherent fashion, but there's no reason that we can't just make a electronic equivalent of a neuron that wouldn't follow the natural processes reasonably well.
I just know that if I am offered the chance to create a drastically flawed upload using one of these methods, I would watch the output very closely. I just know that these would run more efficiently than an actual brain simulation. I'm okay with that because I can't guarantee that computers capable of that are going to exist for a very long time. I am also hesitant to embrace that methodology because simulating reality is inherently less efficient than using approximating functions such as the ones I described before, as running an emulation of a Turing machine on a Turing machine tends to just add a constant value multiplier to time and space complexity while offering very few benefits. So what that they might not have the same qualia? Just so long as the output for a given input approximates the output I would give in that situation, I.E. I can predict its actions and it can predict mine(Thus the Upload and I are very similar), I don't think that it matters terribly much.
As for the organic computers that have been brought up earlier, I trust them even less to hold up to the rigors of interstellar space.
Organic computers will fail in unpredictable and trivially non-correctable ways in an environment rich in hard radiation; among these methods of failure will be cancer. Uncontrollable reproduction is the failure case that I imagine for a biological computer travelling through a radiation-heavy medium such as interstellar space. Alternatively, it would end up with the same sort of neurological damage that certain acute radiation poisonings cause, which is potentially even worse because some of the people affected in such a way by radiation damage ended up becoming very aggressive and incoherent before dying.
As for Jewish Space Slaves, or JSS... I don't see why they'd need to be Jewish. I mean, they can be, but if we're going to embark into the business of Space Slavery(tm), we're going to do it in a 21st century way; free of selection due to religion, ethnicity, or any other conceivable prejudices other than "Are they suitable for the work of space slaves?". This means that while we can be called "Space Slavers"(tm), they won't be able to call us racists or bigots.
Author
Topic: Space Thread (Read 367899 times)