My first brush
with artificial intelligence was during the 1970s. AI was in its infancy and
mostly based upon deducing things from rules, such as “If a person has a headache, then
give him Tylenol.” Software called MYCIN could diagnose bacterial infections. I
suggested adapting this technology to the treatment of human poisoning and
collaborated with the Maryland Poison Center. The software I developed was
called The Interactive Poison Expert for Classification and Control (IPECAC). It
was a fun project, but it really didn’t do very much except allow me to sit at
the Poison Center and listen in on some interesting cases. (Q: “What should I
do? My dog ate a whole jar of Tums!” A: “Take the dog out for Mexican food.”)
Another area of
interest was speech recognition. All kinds of people, all kinds of voices! We
quickly came to the conclusion that this was and would always be impossible.
I don’t have to
tell you how wrong we were. The world has changed. Rapidly! Speech recognition
has become so common that even that most hated human development of all history,
the phone tree, can lead a caller through an interminable conversation while
not taking offense at the perhaps “impolite” language coming in return.
None of this
existed in my youth, and I wonder with some trepidation – no, fear! – what the
world will be like for my grandchildren.
AI technology
continues to advance at an astounding pace. Already, modern AI systems have
beaten chess grandmasters. They can write term papers for dishonest students. And
now, they have begun to produce “thought” patterns unanticipated by their
creators.
Years ago, one of
the inventors of the modern computer, Alan Turing, asked whether computers could
someday “think.” What would this mean? To answer this question, he conceived of
the “Turing Test.” There are two workstations, one connected to a computer and
the other to a person. A user is invited to interact with them any way he
wishes. He can ask questions, play games, or ask for jokes. If the subject is
unable to determine which is the computer or the person, then Turing would say
that the machine “thinks.”
There is a perhaps
apocryphal story about a computer laboratory conducting a Turing test. Various
researchers were unable to determine which was the computer and which was the
person. Finally, they invited a non-technical businessman to try. He sat for a
long time and didn’t do anything. Eventually, they asked, “Aren’t you going to
do anything?”
He replied, “You
said I could do anything I want, so I choose to do nothing!”
A long time
elapsed. And elapsed, and elapsed. Eventually, one of the work stations typed,
“Is anyone there? When does the test start?”
“That,” the
businessman declared, “is the person!”
Well there is
nothing sacrosanct about the Turing Test. Perhaps we should have other criteria
to determine whether the computer thinks, but, to date, the Turing Test has
been passed only in limited experiments.
Will modern AI
eventually produce machines that can be said to think? What about
“consciousness”? Would a thinking machine be a sentient being? What would
happen then? Would there be moral questions? Would it be murder to turn off
such a machine? What would civil libertarians say? These questions are already
being asked. Organizations such as PETA do not distinguish between people and
animals. Is it just a matter of time before AI systems are similarly valued?
What would this
mean for us as humans? Already, there are computer programs that play checkers
perfectly and cannot lose a game. Will this happen with the much more
complicated chess? What about the practice of medicine? A physician friend once
told me that a good clinician should remember every case he has encountered. But
people forget. Computers do not. What about the practice of law? Could a
computer have at its command all the legal cases of the United States courts? What
about thousands of years of halachic writings? A rav I know said that he had a
complex discussion with an AI system about the halachic principle of “mitoch” and it brought forth learned
sources!
Could AI systems
become smarter than any living human? What would we humans then do? Just who
and what are we?
This brings me to
my experience with Rabbi Plony. When I was a graduate student, I volunteered to
visit people in a nursing home. I made some fascinating friends, including one man
who had fought in the Spanish-American War and once shook hands with Teddy
Roosevelt! Well, I don’t know Spanish, but this was “mucho coolo!”
Eventually, I
heard some of the staff wondering if I might be ready for the “sixth floor.”
(Cue the ominous music!) After a while, they asked me if I would be willing to
try it and, not knowing what the “sixth floor” was, I foolishly said “sure.”
Nowadays, such a
place would be called a memory unit. It provided intensive care for people who
suffered from dementias, such as Alzheimer’s disease, or major disabilities,
such as strokes. They introduced me to Rabbi Plony. Looking back, I assume he
had suffered a stroke. He sat in a wheelchair and was unable to move. He could
not converse. The only thing he could do was to say that his name was “Plony,”
which he said over and over again. I was asked to feed him.
I tried to be
cheerful: “Oh look at this nice Jello!” Inside, however, I was in turmoil and
becoming depressed. As he kept on saying “Plony,” I wondered where the rabbi
was. He was a rabbi! – and knew (had
known?) more Torah than I ever would. Where was Rabbi Plony? It did not take a
modern AI system to surpass his ability to interact with the world. Yet he was
a human being with dignity and value far beyond anything a mass of electronic
circuitry could ever be.
Where are we going,
and where was Rabbi Plony? Even in his disabled state we know that even one
second of his life had infinite value. But why? It could not be due to his
abilities in the physical world. It could not be due to his intellect. In his
commentary on Mishlei, the Vilna Gaon
observes that a person has three qualities: Chochma
refers to the information he has learned. Bina
refers to that which he derives through his
understanding. Rabbi Plony did not display either of these. But there is a
third part, and the Gaon writes that this is hidden. It must lie in some place
we cannot observe, somewhere not physical.
Rabbi Plony had a neshama. I realized that I had naively
identified the neshama with
consciousness and personality. This was patently false since Rabbi Plony did
not seem to retain any personality, and I had no idea whether or not he was a
sentient conscious being in the usual sense. Yet he was present, and he was an
infinitely valuable person, a whole world that I could not access. I wondered,
could he? What is it like to be a person experiencing (suffering from)
dementia? Indeed, what is the neshama?
I certainly cannot answer this, but if we
develop AI systems that are even more intelligent than humans, that have
personalities and that “think,” maybe only then will we discover what they are
missing. Maybe only then will we learn what it truly means to be human.
Please, Ribono Shel Olam, let my grandchildren discover what is missing and come
to a better understanding than I have.
