Raúl
Moderator
Moderator
| Posts: 388 |  |
Karma: 6
|
Re:Turing Test - 2007/01/30 13:47
Turing test seems to be a good approach for theoretical discussions, but when it comes to test consciousness for an AI system like CERA [1] it is not applicable. In humans, usual criterion is Accurate Report. This is a behavioral index, as you ask the person to voluntary answer (verbally or any other way) to your questions about his/her perception. Again, this is not easy to extrapolate to other mammals or machines, where these kind of communicative/behavioral capabilities are not available.
Other criteria exist for consciousness in humans and mammals, which are based on brain physiology of consciousness (instead of external behavior). Seth et al. [2] offer 14 distinctive properties of a conscious brain (taking human brain as the reference). If your machine consciousness system is not an imitation of the underlying neural mechanisms in human brain, then these criteria cannot be applied. So, in an artificial system, lacking a thalamocortical core, what criteria for consciousness can we apply?
More oriented to the realm of machine consciouness, Aleksander and Dunmall ([3]) suggest a set of axioms for consciousness that a system should satisfy in order to be considered conscious of the world where it lives.
[1] http://www.conscious-robots.com/en/conscious-machines/machine-consciousness-projects/cera.html
[2] Anil K. Seth, Bernard J. Baars, David B. Edelman. Criteria for consciousness in humans and other mammals. Consciousness and Cognition 14 (2005). Pp. 119-139.
[3] Aleksander, I., Dunmall, B. Neccesary first-person axioms of neuroconsciousness. In: IWANN. Lecture Notes of Computer Science (2003). Pp. 630-637.
Raúl Arrabales Moreno. conscious-robots.com/raul |
Forums... 
