A History and Critique of the Field

What many AI researchers do when they say they are doing AI contradicts what some AI researchers say is AI. Surveys of leading AI textbooks demonstrate a lack of a generally accepted historical record. These surveys also show AI researchers as primarily concerned with prescribing ideal mathematical behaviors into computers -accounting for 987 of 996 (or 99%) of the AI references surveyed. The most common expectation of AI concerns constructing machines that behave like humans, yet only 27 of 996, (or 2%) of the AI references surveyed were directly consistent with this description. Both approaches have shortcomings – prescribing superior behavior into machines fails to scale to multiple tasks easily, while on the other hand, modeling human behaviors in machines can give results that are not always correct or fast. The discrepancy between the kind of work conducted in AI and the kind of work expected from AI cripples the ability to measure progress in the field. This research was supported in part by the Laboratory for International Data Privacy in the School of Computer Science at Carnegie Mellon University.