How Not to Compare AI With the Human Brain

It is self learning

It is a computer, a machine. There is no “self” for it to learn with. It cannot learn because it is a machine. The term machine learning is composed of two words that when combined in that order result in a logical contradiction. It is a logical contradiction to say that a machine can learn because the definitions of each term preclude that possibility. If a machine could learn it would no longer be a machine. Moreover, there are many competing theories as to how the human brain/humans and some non human animals learn all of which have some valid claim to “correctness”. Therefore even if it had a “self” (which it most definitely does not) we could not say that it is self-learning since we do not even know how humans and non human animals learn.

It assumes

It assigns various weights to different outcomes depending on its programming. Some of these outcomes are “preferred” because they are assigned higher weights. The only thing it “assumes” are that these outcomes (the ones assigned greater weights) are the “correct” ones. Essentially it is assuming that larger values/weights are better than smaller values/weights. Like a human assuming 2 > 1 because it is a bigger number. A poor analogy as it does nothing ‘like a human’ for it is a machine.

It adapts

Based entirely on its programming it modifies the weights assigned to various outcomes based on the input data given, it then checks these weights against the “desired” or “correct” outcome (as determined by its programming). When additional input data is given it “adapts” by assigning greater weight to the input data that gives the “correct” output. Essentially the values in the equations it is running change. It does not change the values through force of will or process of self realization since has no will or self, the values change because that is what its programming requires. I think it goes without saying that evolution by natural selection which ‘forces’ adaptive responses in living things if they wish to survive and multiply, does not and will never apply to a machine.

It predicts

With a given set of input data it checks the “correctness” of the weights assigned and output data against what a previous input training data set gave for outputs that were deemed “correct”. It then outputs this data for the user. It has no idea of the past or the future, what is prediction of future or what is retelling of history, it simply outputs certain values based on the input data given and its programming. It is only a prediction in the sense that the computer has no idea of the “rightness” or “wrongness” of its output, nor does the user. The machine may calculate and output a probability of rightness assigned by the computer but that has no bearing on its’ actual rightness, nor does it in any way mean that the output is in any sense a prediction except for in the one way I just mentioned.

It finds typical and untypical patterns

Based on the outcomes of the analysis of various training data sets it “knows” which outcomes are normal and which are abnormal. Given any input data set it can compare the outcomes against previous normal and abnormal outcomes to determine if the results are typical or atypical. All of this is completely determined by its programming. It is not “finding” anything, it is only evaluating the rightness of fit of the answer to a math problem. If the fit is good it is deemed typical, if not, atypical.

It analyses and suggests the most valuable decisions for the user

Not being human or self aware or having any sense of self or morals it can have no idea what is or is not valuable to a user. What is valuable to the user is what the user programs the computer to consider valuable by assigning higher weights to variables and outcomes that increase it and lower weights to those that decrease it. It does not suggest anything either, nor is it capable of making suggestions, it only outputs what its programming requires it to output based on the input data given.