The missing G
- PDF / 863,390 Bytes
- 13 Pages / 595.276 x 790.866 pts Page_size
- 38 Downloads / 172 Views
OPEN FORUM
The missing G Erez Firt1 Received: 23 August 2019 / Accepted: 6 January 2020 © Springer-Verlag London Ltd., part of Springer Nature 2020
Abstract Artificial general intelligence (AGI) is not a new notion, but it has certainly been gaining traction in recent years, and academic as well as industry resources are redirected to research in AGI. The main reason for this is that current AI techniques are limited as they are designed to operate in specific problem-domains, following meticulous preparation. These systems cannot operate in an unknown environment or under conditions of uncertainty, reuse knowledge gained in another problem domain, or autonomously learn and understand the problem-domain. We shall call AI systems capable of such feats artificial general intelligent (AGI) systems. The three tasks of this paper are to provide a working definition of the term AGI, examine the “missing G”, i.e., the set of abilities that current AI systems lack and whose implementation will result in a basic AGI system, and consider different approaches, including a hybrid one, to a comprehensive solution for an AGI. Keywords Artificial general intelligence · Learning · Understanding · Creativity · Abductive reasoning
1 Introduction Artificial General Intelligence (AGI) is not a new notion, but it has certainly been gaining traction in the last decade or so. Conferences hosting researchers from different fields (computer and neuro-sciences, logic and mathematics, philosophy etc.) are dedicated to the study of AGI,1 university courses invite expert guest speakers to discuss different aspects related to the concept,2 and academic as well as industry resources are invested in or redirected to research in AGI. One of the main reasons for this is that AI researchers have realized that current AI techniques are limited. To be sure, Deep Learning networks and their variants (e.g., Convolutional Neural Networks) have reached outstanding achievements in various tasks and even surpassed the predictions of many experts; nevertheless, applications of these techniques are considered “narrow” AI. Why? Because these systems are designed to operate in specific problemdomains following meticulous preparation—usually by being fed predefined models and millions of training examples as input, before they can start working—and even then, their accuracy, although averagely high, is not guaranteed. These systems cannot operate in an unknown environment * Erez Firt [email protected] 1
Philosophy Department, University of Haifa, Haifa, Israel
or under conditions of uncertainty. They cannot use knowledge gained in another problem domain. They cannot autonomously learn and produce a model of the world and act accordingly. They cannot understand the problem-domain and realize what model or parameters should be used and extracted from the environment. We shall call AI systems capable of such feats Artificial General Intelligent (AGI) systems, or at least the first phase of AGI systems. The first task of this paper is to provide a working
