User Tools

Site Tools


public:events:agi-summerschool-2012:selmer

«– HUMANOBS.ORG | SummerSchool Main | Schedule Day-by-Day



Selmer Bringsjord: Logic is the Key to Modeling, Simulating, and (Partially) Replicating Person-Level AGI (Artificial General Intelligence)

Invited Lecture II - August 6, 2012

Core arguments at a glance

1. Personhood includes a set of capacities that exceed those seen in mere animals, and that serve as targets for AGI research and development.

2. The capacities constitutive of personhood are beyond the reach of AGI research that doesn't deploy the tools and techniques of computational formal logic.

3. Point 2 is efficiently appreciated by reflecting upon the mathematical cognition of human persons. The reflection has the side-effect of revealing that the greatest AI achievement to date, IBM's Watson, is a mere animal — but enhancement via formal logic could move the system in the direction of person-level AGI.

Abstract

I suppose there is a sense in which an eagle is general-intelligent. It knows how to fly in remarkably adept fashion, can spot and snatch prey with astounding dexterity, and so on. But I'm not concerned with eagles; in general, I'm not concerned with mere animals. I'm concerned with persons. We are persons: (A1) we can communicate in highly expressive natural and formal languages; (A2) we have and exploit at least fifth-order beliefs (I believe that some of you may believe that I believe that Gödel believed that his mother believed that God exists), whereas eagles can at most muster second-order ones; (A3) we have subjective consciousness (there is something it's like for us to e.g. witness the glory of an Icelandic geyser for the very first time); (A4) we are capable of abstract, formal thought about infinitary objects entirely removed from the hum-drum physical world; etc. But personhood isn't restricted to the case of homo sapiens sapiens. The universe may well contain non-human persons (e.g., aliens, God, angels, etc.).

After briefly defending, contra Darwin, the huge, qualitative divide between mere animals and persons, I explain that while (A3) is beyond the reach of any third-person framework (which implies that ultimately, at best, the AGI program will produce zombies), whether logic-based or not, (A1), (A2), and (A4) can't be computationally modeled, simulated, and (partially) replicated unless the tools and techniques of formal logic are deployed. This explanation is grounded in particular examples involving mathematical cognition. I include analysis showing that study of person-level mathematical cognition reveals that the major success stories of AI to this point, Deep Blue and Watson, are merely animalistic — though in the case of the latter, enhancement via formal logic would yield a system that at least moves closer to the powers of personhood.

Readings For This Presentation

From Our Required Readings List

  • Bringsjord, S. (2008) Declarative/Logic-Based Computational Cognitive Modeling. In R. Sun (ed. ), The Cambridge Handbook of Computational Psychology (Cambridge, UK: Cambridge University Press), pp. 127–169. PDF
  • Bringsjord, S. & Licato, J. (forthcoming) Psychometric Artificial General Intelligence: The Piaget-MacGuyver Room. In Theoretical Foundations of Artificial General Intelligence, edited by P. Wang and B. Goertzel (Atlantis Press). PDF
public/events/agi-summerschool-2012/selmer.txt · Last modified: 2012/08/13 09:13 by thorisson