In considering the question "can machines think,"
there two possible interpretations:
Weak AI: Can machines be made to act as if they
were intelligent?
Strong AI: Do machines that act intelligently have
real conscious minds?
A person can believe in neither, both, or either one of
these propositions.
The question of intelligence and what it means to act
or be intelligent has been debated for over 2000 years
without resolution.
We consider these questions now because unlike other fields, artificial intelligence has generated a thriving industry devoted to proving that it is impossible. Examples:
-What computers Can't Do: A Critique of Artificial Reason by H. Dreyfus
-What computers Can't Do: The Limits of Artificial Intelligence by H. Dreyfus
-What computers Still Can't Do: A Critique of Artificial Reason by H. Dreyfus
-Minds, Brains and Science by J. Searle
-The Rediscovery of the Mind by J. Searle
Although few AI researchers believe that anything
significant hangs on the outcome of the debate over
these questions, popular opinion regarding this debate
may seriously effect public acceptance and support of
the field.
Although he considered the question to be ill-defined, he did address the following objections to the possibility of thinking machines:
1. the theological objection
2. the "head in the sand" objection
3. the mathematical objection
4. the argument from consciousness
5. arguments from various disabilities
6. Lady Lovelace's objection
7. Arguments from continuity in the nervous system
8. arguments from the informality of behaviour
9. arguments from extrasensory perception
We will consider points 1, 4 and 9 when discussing
strong AI. Below we consider the remaining points as
they relate to the possibility of achieving intelligent
behavior---weak AI
Argument: the consequences of thinking machines would be too dreadful to allow the development of such technology.
Refutation: He offers consolation
Argument: There are known limits to the power of computing machines: Godel's incompleteness theorem, and the halting problem.
Refutation: There is no proof that humans do not
have similar limitations.
Argument: There are things that a machine can never do, such as: be kind, be resourceful, be beautiful, have a sense of humor, tell right from wrong, fall in love, be subject to its own thoughts, enjoy strawberries
Refutation: Such scepticism is based on past
experience which may not lead to reliable
predictions about the future.
Argument: A computing machine has no ability to originate anything. It can do only what we know how to order it to do.
Refutation: machines that learn, e.g., Samuel's
checker-playing program.
Argument: Computing machines are digital and the nervous system is continuous.
Refutation: The basic argument may be in
question now. Different systems may have the
same functionality.
Argument: Human behavior is far too complex to be captured by any simple set of rules.
Refutation: We can not convince ourselves of the
absence of such mechanism nor can we identify
the alternative mechanism for initiating and
controlling behavior.
Turing addressed the question as follows:
Argument: To be intelligent a machine must be aware of its own mental state, it must be conscious.
Refutation: In ordinary life we never have any evidence about the internal mental state of other humans so we can not know that anyone else is conscious.
The book presents this debate in terms of two well known thought experiments:
1. The Chinese Room experiment
2. The brain Prosthesis experiment
According to Searle running the right program does
not necessarily generate understanding.
The thought experiment postulates the existence of a
room that speaks Chinese; it receives input written in
Chinese and outputs written Chinese statements. We
are to assume that the system works perfectly.
The system inside of the room consists of:
The argument: We have a system that acts
intelligently, but there is no understanding of Chinese
possible in that system. The person in the room does
not understand Chinese, the book and the stacks of
papers, being just pieces of paper, can not understand
Chinese, therefore, there is no understanding of
Chinese going on.
Refutation: Although the components of the system do
not understand Chinese it does not follow that the
entire system lacks this capability. If you were to ask
the room if it understood Chinese it would answer yes.
Rebuttal: Consciousness is an emergent property of
appropriately arranged systems of neurons. Here the
claim (without a proof) is that consciousness can arise
only from neural material. A position of "biological
naturalism."
This claim leads naturally to the question: Which
properties of neurons are essential to consciousness
and which are merely incidental. This question is
addressed in the next thought experiment.
The experiment: Gradually all of the neurons in a
persons brain are replaced with electronic devices,
then the process is reversed and the person is returned
to their natural state.
The Big Question: Would consciousness vanish when
the neurons are replaced?
Two Possible conclusions:
1. The causal mechanisms involved in consciousness are still present in the electronic version which therefore there is conscious.
2. The conscious mental events in the normal brain have no causal connection to the subject's behavior and are missing from the electronic brain.