Must-Read: Luciano Floridi: A fallacy that will hinder advances in artificial intelligence
Must-Read: There are many inconsistent and wildly different definitions of “artificial intelligence”. Here are two useful ones:
- Building systems that humans can easily and successful interact with by acting as if they are a human-level intelligence (within their domain).
- Building systems that behave in ways “that would be called ‘intelligent’ if a human were so behaving.”
Luciano Floridi says smart things about the second:
Luciano Floridi: A fallacy that will hinder advances in artificial intelligence: “The best definition of AI was written in 1955 by US computer scientist John McCarthy and colleagues… https://www.ft.com/content/ee996846-4626-11e7-8d27-59b4dd6296b8
….The problem, they wrote, “is taken to be that of making a machine behave in ways that would be called intelligent if a human were so behaving”…. The definition does not say the artificial agent is intelligent, but that a human would have to be to achieve the same goal…. The whole and only point is to perform a task such that the outcome is as good as, or better than, what human intelligence would have been able to achieve. “How” is not in question, only the outcome. This is why AI is not about reproducing but replacing human intelligence. A dishwasher does not clean the dishes as I do. But in the end its clean dishes are indistinguishable from mine—indeed, they may be cleaner…. AI is the continuation of intelligence by other means…