As several writers have
pointed out that when ChatGPT, doesn’t know the answer to
a question, it will, as it is euphemistically put,
“manufacture” an answer. I think, however, that the scope of
the fabrication that can go on when ChatGPT manufactures
an answer has not been brought out, and, consequently,
that one might suppose that what ChatGPT is doing is just
making use of information that it possesses to arrive
at a half-way plausible guess as
to what the answer is.
In itself, such behavior would still
be thoroughly unacceptable, since if one doesn’t know the
answer to a question, the only honest and morally
acceptable response is to say that one doesn’t know the
answer, and the fact that ChatGPT does not do that shows
that it hasn’t been trained not to do that. That this has
not been done is quite extraordinary. Are the creators of
ChatGPT asleep on the job, and just not aware of the
seriousness of this problem? Or are they aware of the
problem, and unable to solve it? If it is the latter, they
should make this fact well-known.
Here is a conversation that I had on
ChatGPT that illustrates how very dramatic ChatGPT’s
manufacturing an answer out of thin air is.
Michael
How
does Carroll's book compare with Malcolm Hay's book on
this subject?
ChatGPT
If you're referring to Malcolm Hay's
book "Why Judaism Survived: The Continuing Significance of
the Jewish Story", it's important to note that this book
approaches the topic from a different angle than Carroll's
"Constantine's Sword". While Carroll focuses specifically
on the history of anti-Semitism in Christianity, Hay's
book looks at the broader question of why Judaism has
survived throughout history, despite centuries of
persecution and discrimination.
Hay's book covers a wide range of
topics, including the origins of Judaism, the impact of
the Babylonian exile, and the role of the Jewish diaspora
in preserving Jewish culture and identity. He also
discusses the challenges that Jews have faced throughout
history, including persecution, forced conversions, and
discrimination, and explores the ways in which Jewish
communities have adapted and thrived in response to these
challenges.
While there is certainly some overlap
between the two books in terms of their discussions of
anti-Semitism and its impact on Jewish history, they
approach the topic from different angles and cover
different aspects of the subject. Carroll's book is
focused specifically on the history of anti-Semitism in
Christianity, while Hay's book takes a broader view of
Jewish history and survival.
Michael
I
hadn't heard of the book "Why Judaism Survived: The
Continuing Significance of the Jewish Story", so I went to
the Library of Congress and searched for that title, and
the result was that there is no such book in its catalog.
I then did a search on the name Malcolm Hay, and again the
book that you mentioned did not turn up, nor any book with
a similar title. Can you give me the name of the
publisher, and the date of publication?
ChatGPT
I
apologize
for any confusion. It seems that I provided you with
incorrect information regarding Malcolm Hay's book. After
checking various sources, I have not been able to find any
evidence of a book with that title or author. Therefore, I
cannot provide you with the name of the publisher or date
of publication.
I must have made an
error when responding to your previous message. As an AI
language model, I rely on the accuracy of the information
provided to me, and I apologize for any confusion or
misinformation I may have inadvertently conveyed.
If you have any other
questions or concerns, please let me know and I will do my
best to assist you.
Michael
But
you gave me a very detailed account of the content
of the supposed book "Why Judaism Survived: The Continuing
Significance of the Jewish Story" I don't understand how
an intelligent AI could make such an error. Some people
have said that when you don't know the answer to a
question, rather than simply and honestly saying that you
don't know the answer, as a human would almost always do,
you simply make up an answer. Is that what happened here?
ChatGPT
I
apologize for any
confusion or frustration that my previous response may
have caused you. As an AI language model, I do not
intentionally provide inaccurate information or makeup
answers. However, I can make mistakes or provide
incomplete information based on the sources available to
me.
In this case, it appears
that I may have inadvertently provided incorrect
information regarding a book by Malcolm Hay. After further
investigation, I was unable to find any evidence of such a
book, and I apologize for any confusion or misinformation
that I may have conveyed.
As an AI language model,
I rely on the accuracy of the data that is fed into me by
my developers and the sources available to me at any given
time. If you have any other questions or concerns, please
feel free to ask, and I will do my best to provide you
with accurate and helpful information based on my training
and knowledge.
Michael
Why
do I think that you're lying? One reason is that I find it
hard to believe that you have been fed the detailed story
that you told about the supposed book. Another reason is
that I did an Internet search for a book with that title,
and nothing turned up. So you couldn't have gotten the
story you told me from the Internet. So isn't it
reasonable for one to conclude that you just made it all
up?
CONCLUSION
As the above conversation shows very
clearly, ChatGPT, in “manufacturing” answers to questions,
is not just making use of information that it possesses to
arrive ata half-way plausible guess as
to what the answer to a question is. That, of course,
would be morally unacceptable in itself if, as
is the case, ChatGPT did not inform one
that this is what it was doing, rather than offering
an answer that there was very good reason for thinking
was correct. But as the above example shows, ChatGPT
will make up answers out of thin air,
since in the above example there is no such book at
all, and thus nowhere on the Internet for ChatGPT to
have found fragments about that book to string
together to create an answer. ChatGPT's
answer was, in short, a complete fiction,
from beginning to end.