Lecture 16. The Limits of Computing and
Artiﬁcial Intelligence (cont)
Informal and unedited notes, not for distribution. (c) Z. Stachniak, 2011-2012.
Note: in cases I were unable to ﬁnd the primary source of an image or determine
whether or not an image is copyrighted, I have speciﬁed the source as ”unknown”. I
will provide full information about images and/or obtain reproduction rights when
such information is available to me.
Modern AI does not speak about robots that often. Robots and androids
have been subsumed by a more general class of AI entities called AI agents.
These agents act in diverse environments from real terrains (e.g. a real oﬃce
or Martian terrain) to virtual realities and Internet (e.g. game AI characters
or negotiating agents for e-commerce).
An AI agent can be a physical device or software (or combination of both)
that uses knowledge (represented and stored in the so-called knowledge base)
and speciﬁc knowledge-based methods for solving problems in a selected do-
main. AI agents ”act” by performing actions that modify their environments.
1 Examples: Here are a few examples of AI agents’ types:
• game playing programs (knowledge base of game playing programs con-
sist of rules of games as well as some search and game strategies);
• natural language processing agents for text translation, automated on-
line assistants, call center assistants, etc. (knowledge base contains,
among other things, information about natural languages as well as
about speciﬁc domains);
• AI-planning systems (knowledge base contains information about the
environment, goals to be achieved, and actions available to the agent);
• AI-based veriﬁcation systems designed to detect speciﬁc errors in soft-
ware and hardware designs (knowledge base contains, among other
things, information about the system being diagnosed for errors);
• AI e-commerce agents, such as stock investors, ﬁnancial traders, prop-
erty managers, and fraud detectors.
In the past, the advancement in AI was greeted with a lot of attention from
popular media. After all, these robots, androids, and other types of artiﬁcial
”big brains” were to change our social and economic destiny.
However, after AI shifted its focus away from the Turing test, the disci-
pline became less visible to non-experts. Successful AI solutions have not
been encapsulated in a body of a single android but embedded into myriad
of ”non-AI” products ranging from refrigerators and elevators, to cars, air-
planes, medical, scientiﬁc and military equipment. AI solutions are currently
employed in business, education, arts, and research, in design and manufac-
turing, and, of course, in toys and games. In many cases, the users of the
AI-based technologies are not even aware of the AI component embedded in
their cars and every day products.
The best of AI
It is not possible to list all signiﬁcant (or popular!) AI ideas, achievements,
and applications that have been proposed or developed since the 1960s. One
could insist on stressing great successes of AI in areas such as health, com-
munications, and space explorations. One could point out that AI ”bankers”
2 do quite well in trading and fraud detection, that a damaged plane can self-
diagnose itself and, then, plan and take recovery actions to allow for safe
landing. Or that a spacecraft can autonomously navigate itself, set mission
goals, can make plans to achieve them even when a spacecraft is not fully
Instead, let me present a short list of my AI favourites, some from the past
and some current. The list is arranged in no particular order.
# 1. The Davis-Putnam-Loveland-Logeman algorithm (1960, 1962) and the
Resolution Principle (A. Robinson 1963) laid the foundations of automated
reasoning – a core concept in AI.
3 # 2. Shakey the robot was designed and experimented with at Stanford
Research Institute from 1966 through 1972. The work on Shakey initiated
the AI-based robotics research (learning, planning, AI-oriented programming
languages). Shakey and other similar robots caught the public’s imagination
very fast (articles in the New York Times( 1968), National Geographic Mag-
azine (1970)). At present, Shakey resides at the Computer History Museum
Fig. 1. Shakey the robot at Computer History Museum.
4 #3. ELIZA was an early conversational software designed at MIT by Joseph
Weizenbaum in the mid-1960s. A computer under ELIZA’s control was able
to converse with a human on variety of subjects using mostly word matching
to produce responses. The program was immensely popular; it was even
used for psychotherapy (under the name DOCTOR). Some claimed that it
was the ﬁrst program to passed a ”restricted form” of the Turing test.
Fig. 2. The front page of ELIZA manual for Radio Shack TRS-80 computer (1979).
5 Fig. 3. The last page of ELIZA manual produced by Radio Shack for its TRS-80 computer
6 As Weizbenbaum recollected later:
I was startled to see how quickly and how very deeply people be-
came emotionally involved with the computer and how unequiv-
ocally they anthropomorphized it. Once my secretary, who had
watched me work on the program for many months and therefore
surely knew it to be merely a computer program, start