Naslov
Nazi sex robots: moral reasoning guided by computational complexities

Autori
Šekrst, Kristina

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni

Skup
Zagreb Applied Ethics Conference: Ethics of Robotics and Artificial Intelligence

Mjesto i datum
Zagreb, Hrvatska, 05.06.2017. - 07.06.2017

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Domaća recenzija

Ključne riječi
computational complexity, machine ethics, artificial moral agents, machine learning, moral reasoning, deep learning

Sažetak
Computational complexity classifies computer science problems according to their inherent difficulty, i.e. resources needed to come to a solution. Well-known cases are P and NP problems, former are solvable in polynomial time, while the latter are only verifiable in polynomial time, since the time required increases at least exponentially with the size of the problem. In the field of artificial intelligence, the most difficult problems are AI-complete, and they reflect a similar status as NP-complete problems, being difficult or probably impossible to solve by standard algorithms. AI-complete problems include computer vision-related issues, natural language understanding, and similar real-life reasoning. Could an artificial intelligent agent ever develop a sense of ethics and act accordingly? Current AI systems encounter the issue of lacking common-sense knowledge of the situation, while human beings have background experiences to recognize unusual situations and adjust with ease. Ethical conundrums and different categories of various issues, being presented as real-life situations or as a problem for natural language processing, seem to belong to AI-complete problems. Thus, their resolution depends on the resolvement of the greatest computational issue: can NP-complete problems be reduced to P problems? The modern techniques of neural networks and deep learning can only give a limited notion of understanding, which is often prone to corpus- related problems, again derived from human experiences. Hence, if it is shown that higher- rank problems cannot be reduced to quickly solvable ones, even a marginal grasp of an ethical issue in question still cannot be possible by an artificial intelligent agent.

Izvorni jezik
Engleski

Znanstvena područja
Računarstvo, Filozofija

POVEZANOST RADA