works
Carl Shulman and Anders Sandberg Implications of a software-limited singularity incollection Anumber of prominent artificial intelligence (AI) researchers and commentators (Moravec 1999a; Solomonoff 1985; Vinge 1993) have presented versions of the following argument: 1. Continued exponential improvement in computer hardware will deliver inexpensive processing power exceeding that of the human brain within the next several decades. 2. If human-level processing power were inexpensive, then the software for AI with broadly human-level (and then superhuman) cognitive capacities would probably be developed within two decades thereafter. Therefore, 3. There will probably be human-level AI before 2060.

Implications of a software-limited singularity

Carl Shulman and Anders Sandberg

ECAP10: VIII European Conference on Computing and Philosophy, Munich, 2010, pp. 463–470

Abstract

Anumber of prominent artificial intelligence (AI) researchers and commentators (Moravec 1999a; Solomonoff 1985; Vinge 1993) have presented versions of the following argument: 1. Continued exponential improvement in computer hardware will deliver inexpensive processing power exceeding that of the human brain within the next several decades. 2. If human-level processing power were inexpensive, then the software for AI with broadly human-level (and then superhuman) cognitive capacities would probably be developed within two decades thereafter. Therefore, 3. There will probably be human-level AI before 2060.

PDF

First page of PDF