10 AI IN THE OPERATIONAL MANAGEMENT OF LARGE-SCALE IT SYSTEMS
2. Basic AI concepts
2.1. A brief overview of AI methods and applications
Although AI has permeated many aspects of our daily lives by chang ing the way we work, travel, shop or find
partners it did so apparently despite of users lack of understanding of what was going o n inside the proverbial
black box. In spite of the general suspicion towards AI widespread among non-specialist and techno-pessimist
audiences, in the opposing camp there is a strong belief in the promise of AI to help build a better world. It can
be argued that AI, just like any technology before it, includ ing electricity, genetic engineering or the internet,
can be harnessed for the benefit of humanity at large, assuming that the necessary preventive m easures have
been put in place constraining the space for opportunity to use technology against humanity .
First, it is important to distinguis h between narrow or specific AI and general AI, s ometimes also referred to as
machine superintelligence artificial intelligence equalling or surpassing the level of hum an intelligence.
Narrow AI can be defined as the application of AI in a narrowly defined domain, performing a strictly
defined functions or tasks. All currently available AI technologies can be defined as narrow AI, given
that they can only carry out specific tasks, a lbeit often better than humans (e.g. solving complex
mathematical problems, sorting large quantities of data, or playing chess or Go).
General AI or machine sup erintelligence can be defined as artificial intelligence that is equal to or
surpasses human intelligence in all domains of human activity.
Although science fiction has been concerned with the effects of general AI for a few decades already, most
scientists believe that despite the major progress in artificial intelligence, such as the development of art ificial
neural networks and deep learning, we a re still at least a few decades away from the development of general
purpose AI19. As argued elsewhere, the fact that an artificially intelligent system has re portedly passed the
Turing test20 does not mean that AI has reached the level of human intelligence; it only points to the deficiencies
of the test itself21.
Second, it is important to distinguish between the different approaches or stages of evolution o f AI, starting
with symbolic AI, to machine learning, to artificial neural networks and deep learning. Symbolic or classical AI
was the dominant paradigm of AI during the three decades from the mid-1950s to the late 1980s. At the core of
these systems is a computer programme based on exper t knowledge. Such systems follow step -by-step if-
then-else procedures in order to produce the intended ou tcomes In the case of relatively simple processes,
expert systems can rely on true/false values to provide inputs for a utomatic decision making. In more complex
cases, such as medical diagnoses or stock market trading platforms, expert systems rely on fuzzy logic to deal
with a larger number of variables and values where the d ecision is made on the basis of a truth value within
the range of 0 to 122 The huma n in the loop principle applies to such s ystems by default since the systems
essentially follow human decision-making procedures. Since such systems are developed using relatively simple
code, the decision-making processes in such systems can be easily subjected to human audit.
19 National Science and Technology Council (2016) Preparing for the future of a rtificial intelligence. Washington, DC: Executive Office of the President.
Available online: https://obamawhitehouse.archives.gov/sites/default/files/whitehouse_files/ microsites/ostp/NSTC/preparing_for_the_future_of_ai.pdf
20 Turing test is a test of a machine s ability to exhibit intelligent be haviour equivalent to or indistinguishable from that of a human being, n amed after the
English computer scientist Alan Turing.
21 The Eugene Goostman chatbot developed by a team of Russian and Ukrainian p rogrammers passed the test in 2014 by tricking non-expert judges using a
number of techniques, such as an encoded personality (of a 13-year-old Ukra inian boy with imperfect English), as well as using jokes or changing the subject
of conversation (akin to experienced politicians) whenever the question asked did not correspond with a pre-programmed response. This announcement was
received with widely shared scepticism by AI experts. Mitchell (2019) Artificial Intelligence. A Guide for Thinking Humans. New York: Farrar, Straus and