Why Artificial Intelligence is a Piece of Software and not Sentient
Here is a typical description of what happens with software:
“Consider encountering a computer connected to an input wire and an output wire. If we know what algorithm the computer implements, then there are two distinct ways to be uncertain about the output. We could be uncertain about the input — maybe it’s determined by a coin toss we didn’t see. Alternatively, we could be uncertain because we haven’t had the time to reason out what the program does — perhaps it computes the parity of the 87,653rd digit in the decimal expansion of π, and we don’t personally know whether it’s even or odd. “
Found on arxiv: https://arxiv.org/pdf/1609.03543.pdf
But in the biological sphere, we have a third uncertainty. We do not know whether the computer has changed in the meantime. We disregard this in software computation, even though even modern computers have faults in their hardware and some faults go on and off over time and distort the results. But in biology, it is the rule that the computer itself changes, and it is not a fault or a bug, it is a feature that we use and rely on.
“The first type of uncertainty is about empirical facts. No amount of thinking in isolation will tell us whether the coin came up heads. To resolve empirical uncertainty we must observe the coin, and then Bayes’ theorem gives a principled account of how to update our beliefs. The second type of uncertainty is about a logical fact. “ Found on arxiv: https://arxiv.org/pdf/1609.03543.pdf
But in biology there is no logical algorithm that the system invariably computes. There is a third type of uncertainty which is part of the system: We need to anticipate the type of change in the computer, in the system, that may occur and draw our conclusions with hedges.
Or, as they say, that’s life.