Building Safer AGI by introducing Artificial Stupidity

Artificial Intelligence (AI) achieved super-human performance in a broad variety of domains. We say that an AI is made Artificially Stupid on a task when some limitations are deliberately introduced to match a human's ability to do the task. An Artificial General Intelligence (AGI) can be made safer by limiting its computing power and memory, or by introducing Artificial Stupidity on certain tasks. We survey human intellectual limits and give recommendations for which limits to implement in order to build a safe AGI.

[1]  Gary Roberts,et al.  Parsing the Turing Test: Philosophical and Methodological Issues in the Quest for the Thinking Computer , 2008 .

[2]  Daniel Nettle,et al.  The Paranoid Optimist: An Integrative Evolutionary Model of Cognitive Biases , 2006, Personality and social psychology review : an official journal of the Society for Personality and Social Psychology, Inc.

[3]  A. M. Turing,et al.  Computing Machinery and Intelligence , 1950, The Philosophy of Artificial Intelligence.

[4]  R. Marois,et al.  Capacity limits of information processing in the brain , 2005, Trends in Cognitive Sciences.

[5]  G. A. Miller.  THE PSYCHOLOGICAL REVIEW THE MAGICAL NUMBER SEVEN, PLUS OR MINUS TWO: SOME LIMITS ON OUR CAPACITY FOR PROCESSING INFORMATION 1 , 1956 .

[6]  Nick Bostrom,et al.  Whole Brain Emulation , 2008 .

[7]  H. Simon,et al.  A Behavioral Model of Rational Choice , 1955 .

[8]  Nick Bostrom,et al.  Superintelligence: Paths, Dangers, Strategies , 2014 .

[9]  A. Tversky,et al.  Extensional versus intuitive reasoning: the conjunction fallacy in probability judgment , 1983 .

[10]  Demis Hassabis,et al.  Mastering the game of Go with deep neural networks and tree search , 2016, Nature.

[11]  Stuart M. Shieber,et al.  Lessons from a restricted Turing test , 1994, CACM.

[12]  Joseph Weizenbaum,et al.  and Machine , 1977 .

[13]  S. Baum.  A Survey of Artificial General Intelligence Projects for Ethics, Risk, and Policy , 2017 .

[14]  K. Fitzpatrick,et al.  Delivering Cognitive Behavior Therapy to Young Adults With Symptoms of Depression and Anxiety Using a Fully Automated Conversational Agent (Woebot): A Randomized Controlled Trial , 2017, JMIR mental health.

[15]  L. Ross.  The Intuitive Psychologist And His Shortcomings: Distortions in the Attribution Process1 , 1977 .

[16]  John Tooby,et al.  Better than rational: Evolutionary psychology and the invisible hand , 1994 .

[17]  M. Haselton,et al.  The Evolution of Cognitive Bias , 2015 .

[18]  G. Gigerenzer.  Ecological intelligence: An adaptation for frequencies , 1997 .

[19]  A. Tversky,et al.  Judgment under Uncertainty: Heuristics and Biases , 1974, Science.

[20]  Yoshua Bengio,et al.  Gradient-based learning applied to document recognition , 1998, Proc. IEEE.

[21]  Marcus Hutter,et al.  Universal Artificial Intelligence: Sequential Decisions Based on Algorithmic Probability (Texts in Theoretical Computer Science. An EATCS Series) , 2006 .

[22]  Santosh S. Vempala,et al.  The complexity of human computation via a concrete model with an application to passwords , 2020, Proceedings of the National Academy of Sciences.

[23]  V. A. Harris,et al.  The Attribution of Attitudes , 1967 .

[24]  W. E. Hick.  Quarterly Journal of Experimental Psychology , 1948, Nature.

[25]  R. Williams,et al.  The control of neuron number. , 1988, Annual review of neuroscience.

[26]  Yang Cai,et al.  How Many Pixels Do We Need to See Things? , 2003, International Conference on Computational Science.

[27]  Roman V. Yampolskiy,et al.  Turing Test as a Defining Feature of AI-Completeness , 2013, Artificial Intelligence, Evolutionary Computing and Metaheuristics.

[28]  S. Thorpe,et al.  How parallel is visual processing in the ventral pathway? , 2004, Trends in Cognitive Sciences.