Rethinking, Reworking and Revolutionising the Turing Test
- PDF / 852,677 Bytes
- 6 Pages / 439.37 x 666.142 pts Page_size
- 45 Downloads / 176 Views
Rethinking, Reworking and Revolutionising the Turing Test Nicola Damassino1 · Nicholas Novelli1 Published online: 8 December 2020 © The Author(s), under exclusive licence to Springer Nature B.V. part of Springer Nature 2020
1 Introduction We are pleased to present this special issue of Minds and Machines on “Rethinking, Reworking and Revolutionising the Turing Test,” which showcases the latest scholarship on Turing’s Test. We hope it will promote a broader understanding of its continuing relevance today, and perhaps encourage other scholars to engage further with the issues it presents. In the last 70 years, there has been a great deal of thought and writing about the topic, and it continues to have relevance to this day–in fact, recent developments have made understanding of and debates about the Turing Test (TT) all the more vital. The articles in this collection take these discussions in innovative and exciting directions. The common understanding of the TT is that if a computer, through conversation, can fool a judge into thinking it is human, it counts as intelligent. However, the precise details of the test are crucial, as they reveal that the experimental design of the TT is not so simplistic and involves quantifiable comparisons and benchmarks. This should be kept in mind, especially now that our daily lives are increasingly filled with interactions with more and more sophisticated natural language processing (NLP) systems and artificial intelligences (AIs), such as digital assistants like Alexa and Siri, or tech support and troubleshooting bots. Those are all designed to feel as human as possible when we talk to them. This would have been barely imaginable not only in Turing’s day, back when he wrote his original paper in 1950, but even decades later when thinkers from different disciplines were sparking new discussions about the test. If such things were the realm of science fiction until not long ago, far removed from actual experience, today a machine that can understand and respond to our normal spoken sentences is downright mundane. Yet despite the superficial sophistication of these programs, we intuitively know they do not possess genuine intelligence. Without a proper understanding of the true essence of the TT, the TT would be refuted as a * Nicola Damassino [email protected] Nicholas Novelli [email protected] 1
University of Edinburgh, PPLS 3 Charles St, Edinburgh EH8 9AD, UK
13
Vol.:(0123456789)
464
N. Damassino, N. Novelli
useful test. One goal of this work is to investigate the true demands and the adequate experimental design of the TT. There are different principled ways in which a properly formulated TT could separate our current, relatively simple chatbots and AIs from the kind of truly intelligent machines that remain the stuff of sci-fi. Some of these ways will be explored in this collection. Of more concern than AIs intended to imitate humans for the sake of comfort and ease-of-use, are malicious AIs, intended to imitate humans for the purpose of deception.
Data Loading...
