Several attempts have been made during the recent to-do over runaway AI to calm the fears of the ordinary layperson—unsettled by imaginings of anthropomorphic ghosts in the machine—by stressing that computers, being deterministic, are incapable of free will and therefore cannot have motives, evil or otherwise. This simplistic misdirection ignores the actual problem.
Chaos theory demonstrates amply that purely deterministic processes can have unpredictable outcomes. A large part of AI research is focused on developing self-modifying automata capable of rewriting their own code on the fly. These dynamically evolving programs, creatively adapting to challenges encountered in their “world”, embody a phenomenon considerably more complex and unpredictable than that of conventional hardcoded programs.
There is nothing intrinsic to this mechanism that requires consciousness. Given the proper setting, a sufficiently sophisticated self-modifying algorithm left to its own devices might unthinkingly ignite an explosive growth in its own complexity that far exceeds the scope of its initial parameters. Anthropomorphic or not, this is what we have to fear.
People are also reading…
Robert Gavlak
Midtown
Disclaimer: As submitted to the Arizona Daily Star.
Follow these steps to easily submit a letter to the editor or guest opinion to the Arizona Daily Star.
