On the verisimilitude of artificial intelligence

Logique Et Analyse- 190 (189):323-350 (2005)
  Copy   BIBTEX


This paper investigates how the simulation of intelligence, an activity that has been considered the notional task of Artificial Intelligence, does not comprise its duplication. Briefly touching on the distinction between conceivability and possibility, and commenting on Ryan’s approach to fiction in terms of the interplay between possible worlds and her principle of minimal departure, we specify verisimilitude in Artificial Intelligence as the accurate resemblance of intelligence by its simulation and, from this characterization, claim the metaphysical impossibility of duplicating intelligence, as neither verisimilarly nor convincingly simulating intelligence involves its duplication. To this end, we argue by a representative case of simulation that, albeit conceivable, Turing’s test for machine intelligence wrongly equates the occurrence of indistinguishable intelligence performance to intelligence duplication, which is grounded in a prima facie conceivable but metaphysically impossible view that separates intelligence from its origin. Finally, we establish the following criterion for evaluating simulation in Artificial Intelligence: simulations succeed in AI if and only if they are able to epistemically persuade human beings that intelligence has been duplicated, that is, if and only if verisimilar simulations can convincingly minimally depart from actual intelligence.

Author's Profile

Rodrigo González
University of Chile


Added to PP

254 (#38,788)

6 months
26 (#57,221)

Historical graph of downloads since first upload
This graph includes both downloads from PhilArchive and clicks on external links on PhilPapers.
How can I increase my downloads?