[Paleopsych] Ubiquity: Reflections on the Limits of Artificial Intelligence Security

Premise Checker checker at panix.com
Tue Jan 25 15:29:38 UTC 2005


Reflections on the Limits of Artificial Intelligence Security
http://www.acm.org/ubiquity/views/v5i38_tugui.html

                   Ubiquity - The ACM IT Magazine and Forum

           Reflections on the Limits of Artificial Intelligence[i]

       Nature is very simple and efficient in everything she makes. We,
                          humans, complicate things.

    By Alexandru Tugui[ii]

    Abstract: Nature is very simple and efficient in everything she makes,
    and is extremely obvious. We humans like to simulate in an extremely
    complicated manner what exists quite simply in nature, and what we
    succeed in simulating falls in the category of artificial
    intelligence. Artificial intelligence has limits of scope, but they
    fade away when compared with the performances of natural intelligence.
    In this study, we undertake to outline some limits of artificial
    intelligence compared to natural intelligence and some clear-cut
    differences that exist between the two.

    Keywords: intelligence, limits, artificial, robots, digit, law of
    entropy, ABAC.

    Man has always had nature and environment as models for his various
    achievements, and using those models he succeeded in making machines,
    tools, and robots with impressive performances. It's been speculated
    that in the next 40 - 45 years computers could reach the performance
    of the human brain, and that achievements in the field of artificial
    intelligence will be quite amazing, even to the extent of making
    synthetic workers (humanoid robots) with performances very much like a
    human's. Could all this be really possible? We shall live (some of us
    by our offsprings) and ... we shall see!

    Communication leads to intelligence

    The basic idea of the living world consists in transmitting to future
    generations what is considered important for survival. Over time, a
    collective memory was created. The modality of transmitting
    information was complex: from signs to body language, from drawings to
    speaking as such, from chopped stone to magnetic support.
    Communication eventually reached various performances:

    1. From simple to simple, for instance: speaking - speaking, seeing -
    speaking, speaking - seeing;

    2. from simple to complicated, for instance: speaking - representing,
    seeing - representing;

    3. from complicated to complicated, for instance: representing -
    representing.

    Note: Representing is a codification approach according to a certain
    algorithm.

    The last two variants of communication always assumed a certain
    technology by means of which communication as such could be made:
    seeing, speaking, writing, reading. Under such circumstances, we
    should notice the fact that whereas some people had direct access to
    this collective memory, other people had an indirect access by
    traditions and habits.

    It is common knowledge that an immediate connection between natural
    intelligence and the modality of transmitting information and
    knowledge among the members of a community is made directly. Consider,
    for instance, the situation of children raised in the wild, who take
    over the collective memory of the animals with which they live.

    In other words, from the beginnings of mankind to the present, people
    tackled the issue of sharing information among the members of the same
    community or of different communities, of the same times or of
    different times. A first stage consisted in the communication based on
    simple drawings. This communication modality was simple and accessible
    to all people without needing additional training, because everything
    was visual!

    Then the period of using certain symbols followed, symbols that are
    hard to understand by those who are unskilled, which limited the
    access to the transmitted/communicated information. This stage also
    comprises the period of alphabet use.

    The information period assumed the complication of the modality of
    transmitting data, information, knowledge.

    At first sight, everything seems very simple, but data digitizing
    assumes a set of operations that would eventually lead to the
    representation as strings of 0 and 1. We show this with the following
    example: we find by sheer accident a magnetic support on which data
    and information are stored. To know (see) such data and information,
    we have to access them with a specialized peripheral depending on the
    magnetic support, to manage them with a utilitary program, and finally
    to process them with specialized software. All these steps are
    necessary because simply seeing such data and information is of no
    help.

    Considering the above, we may synthetize communication between two
    individuals into one of the following three variants:
    Man (1) - Man (1)
        Man (1) - MMMan (natural) (1)
        Man (1) - intermediary - Man (2)
        Man (1)- peripheral - MMMan (radio, TV, phone, computer
        peripheral) (2)

    Note: MMMan - many men. The index 1 accounts for the fact that the
    communication process is simultaneous, whole index 2 gives us the clue
    that there is a discrepancy between the two moments.

    We think that artificial intelligence must take into account the
    communication modalities and the coupling of the intelligent entities
    to the collective memory of every community.

    When will a computer "grow up"?

    As we consider the evolutionary character of artificial intelligence,
    we naturally wonder When will a computer "grow up"? -- when we could
    speak of an "artificial intelligence" of matter, and have contextual
    procedures that cover most circumstances that occur. In other words,
    we speak of a transmitted intelligence based on limited,
    difficult-to-generalize case studies.

    Anyway, there are situations of denial of natural intelligence -- and
    we have to acknowledge that, once "formatted," the raw matter of
    natural intelligence can hardly ever be recovered. We are speaking of
    children grown in the wild, who lose some of the partial
    characteristics of their natural intelligence -- for example the
    capacity of formulating sentences even if they have a quite rich
    vocabulary.

    People start learning when they are young! They learn from others'
    experience (by the rules transmitted via various modalities); they
    learn from their own experience (by the rules they compile).

    We all accept the idea that natural intelligence is specific to the
    living world, and from this perspective we cannot imagine that
    humankind will ever reach that level of development that would enable
    the simulation of natural intelligence in full detail. We think this
    could be possible only to a certain extent, only if a hybrid system
    between the living cell and the technical system is made -- the
    so-called bio-techno-system. But one should not misunderstand this
    idea! The bio-techno-system does not assume the achievement of a
    technical system having incorporated a sequence of software procedures
    to simulate the biological system, but the coupling of a living
    organism and of a technical system where the informational interaction
    is made by means of the computer system.

    The answer to the question above is the more obvious the closer we are
    to a success in the bio-techno-system field.

    Some limits of artificial intelligence

    We see on various Internet sites posted discussions, courses and
    opinions on the future performance of artificial intelligence
    application fields.

    Nevertheless, the specialists in the field are challenged to create
    equipment and software able to cope with the performance of the human
    brain. There are assessments of time and memory requirements,
    operation speed, ethics regarding how such an artificial intelligence
    system should look and operate. There are even worries that humankind
    will have to face an additional risk if some intelligent informatic
    entities are not restricted in running processes and in making major
    decisions, if they can program themselves (re-writing codes,
    re-compiling), etc.

    Nevertheless, at this point we should take into account certain
    restrictions and limitations pertaining to artificial intelligence
    that we will not succeed in overcoming. These are some of these
    limits:

    1. Artificial intelligence must take into account the law of entropy.
        At this point, the relevant achievements do not take them into
        account and do not succeed in simulating them. In nature, the law
        of entropy leads to the stabilization of any type of system. The
        passage from a high level of entropy to a low one and vice-versa
        consumes energy. Most common movements in nature are the result of
        applying the law of entropy to a given system. We think that by
        the symbiosis between the living cell (living organism) and the
        technical systems, the intelligent control of matter could be
        achieved;
        2. The entire foundation of artificial intelligence is based on
        informatic procedures that mean to circumscribe the intelligent
        behaviour of a human being, although experts never succeeded in
        simulating the behaviour of an ape with an ABAC. As we saw
        previously, the human being has the quality of complicating things
        very much when he knows what he must do but mostly when he does
        not know where he is heading. Therefore, we consider that when the
        goal is not known very well, the human brain both functionally and
        structurally will complicate even further the solution procedures,
        which consumes time and considerable information resources. We
        strongly believe that bio-techno-systems can be a solution to this
        problem;
        3. The two pillars of computer science, "0" and "1" together with
        the truth values "True" and "False" are major borders in
        artificial intelligence. Any intelligent information procedure is
        decomposed eventually in strings of "0" and "1", which leads us to
        the fundamental objection that intelligent machines will never be
        like humans. We have to consider that bio-systems also work with
        intermediary values;
        4. Artificial intelligence is based very much on symbolic logic,
        and has not succeeded in involving so-called affective logic. In
        affective logic, combinations of truth values may lead to
        different evaluations. A possible solution could be obtained by
        using affective computing [1], which undertakes to model affective
        behavior in various situations.

    Conclusion

    We believe that we will make considerable progress in the applicative
    and theoretical fields of artificial intelligence. The limits we
    synthetized are and will be felt for a long time, yet they will
    decrease as new materials and new technologies are discovered. At the
    same time, bio-techno-systems will be solution with a particular
    technological impact on the evolution of artificial intelligence.

    Footnotes

    1. [8]http://www.bartneck.de/link/affective_portal.html

    Bibliography

    1. Bergeron, B. (2002) Dark Ages II. When the Digital Data Die,
    Prentice Hall PTR, New Jersey, 2002

    2. Denning, P. J., Metcalfe, R.M. (eds.) (1997) Beyond Calculation.
    The Next Fifty Years of Computing, Copernicus, Springer-Verlag, New
    York

    3. Mesarovic, M., Pestel, E. (1975) Mankind at the Turning Point: The
    Second Report to the Club of Rome, Reader's Digest Press, New York

    4. Moore, A.D. (1969) Invention, Discovery and Creativity, Anchor
    Books, New York

    5. Tugui, A., Fatu, I. (2004) What is the Globally Information Based
    Society Followed By? in Cyber Society Forum, at
    [9]http://www.wfs.org/04tuguifatu.htm

    6. Tugui, A. (2004) Calm Technologies in a Multimedia World, in
    Ubiquity, ACM, Vol. 5, Issue 4, 17-23 March.

                                     ***

    1.[10]http://www.computergames.ro/forum/archive/index.php/t-54808.html

    2. [11]http://alicebot.sourceforge.net/alice_page.htm

    3. [12]http://www.bartneck.de/link/affective_portal.html

    4. [13]http://www.abelard.org/turing/tur-hi.htm

    Alexandru TUGUI, Ph.D., is a Senior Lecturer at "Al. I. Cuza"
    University, Iasi, Romania

    Source: Ubiquity, Volume 5, Issue 38, December 1 - 7, 2004,
    http://www.acm.org/ubiquity/

References

    7. http://www.acm.org/ubiquity/bk_ex.html
    8. http://www.bartneck.de/link/affective_portal.html
    9. http://www.wfs.org/04tuguifatu.htm
   10. http://www.computergames.ro/forum/archive/index.php/t-54808.html
   11. http://alicebot.sourceforge.net/alice_page.htm
   12. http://www.bartneck.de/link/affective_portal.html
   13. http://www.abelard.org/turing/tur-hi.htm
   14. http://campus.acm.org/forums/ubiquity/messageview.cfm?catid=1&threadid=355



More information about the paleopsych mailing list