What Would It Mean for a Machine to Have a Self?

37 Pages Posted: 15 Sep 2022

See all articles by Julian De Freitas

Julian De Freitas

Harvard University - Business School (HBS)

Ahmet Kaan Uğuralp

Bilkent University

Zeliha Uğuralp

Bilkent University

Laurie Paul

affiliation not provided to SSRN

Joshua B. Tenenbaum

Massachusetts Institute of Technology (MIT) - Department of Brain and Cognitive Sciences

Tomer D. Ullman

Harvard University

Date Written: September 1, 2022

Abstract

What would it mean for autonomous AI agents to have a ‘self’? One proposal for a minimal notion of self is a representation of one’s body spatio-temporally located in the world, with a tag of that representation as the agent taking actions in the world. This turns self-representation into a constructive inference process of self-orienting, and raises a challenging computational problem that any agent must solve continually. Here we construct a series of novel ‘self-finding’ tasks modeled on simple video games—in which players must identify themselves when there are multiple self-candidates—and show through quantitative behavioral testing that humans are near optimal at self-orienting. In contrast, well-known Deep Reinforcement Learning algorithms, which excel at learning much more complex video games, are far from optimal. We suggest that self-orienting allows humans to navigate new settings, and that this is a crucial target for engineers wishing to develop flexible agents.

Suggested Citation

De Freitas, Julian and Uğuralp, Ahmet Kaan and Uğuralp, Zeliha and Paul, Laurie and Tenenbaum, Joshua B. and Ullman, Tomer D., What Would It Mean for a Machine to Have a Self? (September 1, 2022). Harvard Business School Marketing Unit Working Paper No. 23-017, Available at SSRN: https://ssrn.com/abstract=4220026 or http://dx.doi.org/10.2139/ssrn.4220026

Julian De Freitas (Contact Author)

Harvard University - Business School (HBS) ( email )

Soldiers Field Road
Morgan 270C
Boston, MA 02163
United States

Ahmet Kaan Uğuralp

Bilkent University

Bilkent, Ankara 06533
Turkey

Zeliha Uğuralp

Bilkent University

Bilkent, Ankara 06533
Turkey

Laurie Paul

affiliation not provided to SSRN

Joshua B. Tenenbaum

Massachusetts Institute of Technology (MIT) - Department of Brain and Cognitive Sciences ( email )

43 Vassar Street
Cambridge, MA 02139
United States

Tomer D. Ullman

Harvard University

1875 Cambridge Street
Cambridge, MA 02138
United States

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
90
Abstract Views
276
rank
405,503
PlumX Metrics