I had an email from my friend Fred, about ChatGPt, and I offer his excellent commentary, about my play with ChatGPT, just so I can find it! I set the intending to put this in a work, moving forward, the next booklet.
"I see you mentioned me and my "hate" for ChatGPT.
Let me clarify. I don't hate ChatGPT. If it helps you get your thoughts
together I don't object to that.
I do object to imagining that it has some kind of actual wisdom or
intelligence, to giving it credit as a "collaborator" or to otherwise
thinking of it as some kind of entity. And I object strongly to the
business model of both text and image-based generative AI programs that
are all based on ripping off massive amounts of original copyrighted
writing and artwork without permission or compensation to the creators.
I have learned enough about how it works to understand that ChatGPT and
other text-based "large language models" are basically more complex
versions of the predictive text models built into programs like GMail.
In those when you type a word the program guesses the most likely
following words or phrases based on what people commonly type. It is
trained on massive amounts of text input. It has no actual experience or
knowledge of anything other than text, no way of distinguishing between
reliable and unreliable information, and no actual "intelligence" other
than recognizing common linguistic patterns.
It often seems to work like a student plagiarizing a term paper by
copying a source and then going through it with a thesaurus to
substitute a lot of the words with similar words. I've seen many
examples of ChatGPT text that look uncannily like that. Experts have
described it as a "plagiarism machine".
When it is asked to imitate a famous author's writing style, it uses
that author's texts as sources for its "predictive text" function. Thus
it may use similar words and sentence structures, which give its
generated text a superficial resemblance to the author it is mimicking.
It has absolutely no understanding of the meaning behind the author's work.
Because the software works purely by imitating existing texts, without
any underlying understanding of the mental process behind the words and
without any actual experience of the world, it will always produce the
most predictable, the most common, the shallowest ideas about whatever
subject it is asked to address. That is what it is designed to do. If
you are trying to produce reasonable-sounding bland content-free
pabulum, it is the perfect tool for that.
In creative work, we should choose tools that harness our own
creativity, our life experience, our powers of analytical and
imaginative thought, and our organic sense of beauty. AI tools can be
useful - for example, there are photo retouching tools that can replace
a messed-up part of a photograph with imagery that matches the rest of
the photo. That kind of thing can be very useful, replacing hours of
tedious manual work. But in that case it's not doing the creative part
of the work, just the boring part of the work. The problem I see with
using AI in the creative field is that people are letting it do what
should be the actual creative work.
The work that you are doing with ChatGPT does not change my view. I
don't see it producing anything better than what I know you are capable
of writing with your own skills - in fact, quite the opposite. None of
this seems in any way up to the level of some of the writing you've done