Latest / HumanPrint: How to Use AI Without Losing Yourself / Episode 13: Freedom of Creative Process

Transcript

Episode 13: Freedom of Creative Process · HumanPrint: How to Use AI Without Losing Yourself

0:00
Foreign.
0:05
Welcome to Human Print. I am Christine Whitmarsh,
0:08
and if you are open to collaborating with AI but you're not
0:12
interested in disappearing into AI, you're in the right place.
0:16
Each week we will look at one place where your authentic human signal
0:19
is at risk for drift and one way to shore it up.
0:23
Oh, and Human Print is brought to you by Abby AI
0:27
Brand bank intelligence that helps authors, thought leaders and enterprise
0:32
teams train AI on their voice and point of view so written
0:37
content stays unmistakably human, consistent and on brand.
0:41
Learn more@abbey.ink a b
0:45
b I.ink like you write with. Let's begin.
0:52
Hi everybody and welcome back to Human Print.
0:56
All right, over the past few episodes, we've been working through some pretty fundamental
1:00
questions about writing. Like why do you write? How do you sound like
1:04
you? What do you actually want to write about? Those are the last three
1:08
episodes. And now we get to the part that nobody else can really
1:12
decide for you, even though a growing number of technology companies and self appointed
1:16
AI detectives seem very eager to try.
1:19
And that question is, how do you write?
1:23
What happens between the first thought entering your mind and the finished
1:27
piece appearing in front of a reader? Where does your thinking happen?
1:31
What helps you develop it? What interrupts it? What strengthens
1:36
it? And where in your own process do you draw the line between receiving
1:40
useful assistance and handing over something you needed to do yourself?
1:45
That process, all of it, belongs to you.
1:49
Because a writer writes. But a writer also talk,
1:53
might talk, dictate, outline, research,
1:56
interview, rearrange, cut, rewrite,
2:00
stare at the same paragraph for 45 minutes, call someone and explain
2:04
the whole idea out loud. And then return to the paragraph and finally understand what
2:07
they were trying to say. Yeah, I've been there. Some writers begin with a detailed
2:11
outline. Some have to write several meandering pages before they discover their
2:16
subject. Some, like me, need silence. Some need music.
2:19
Some write by hand, some dictate into their phones, some cover a table
2:23
with index cards. And some. Some of them, you know, we open a blank
2:27
document and start working without any idea what the piece is going to be.
2:31
The finished work does not reveal every step
2:35
that produced it. Which brings me to what Substack recently
2:39
rolled out. They recently rolled out Substack, the social media
2:43
for writers platform. In case you're not familiar, they recently introduced a
2:46
feature allowing readers to SC posts, notes, comments and replies
2:51
using the AI detection company Pangram. I don't know if
2:55
that's how you pronounce it. Anyway, the tool then estimates how much
2:59
of the text was human written, AI assisted, or AI generated.
3:05
Writers can add a public statement explaining how I
3:08
make this. So this is my process. This is how I made this piece.
3:12
They can also, what I do, disable detection on an individual piece,
3:16
although readers who attempt to scan it will then see the message AI
3:20
detection unavailable. So those are the options.
3:24
Submit to the scan, explain your process in advance,
3:28
turn off the scanner, and accept whatever somebody decides that means.
3:32
Or deal with the result after a machine has assigned percentages to your
3:36
writing. Yeah, no thanks. And I know the immediate response
3:40
from someone will be, you know, well, if you have nothing to hide, why would
3:44
you care? Well, you know what? Privacy is not evidence of wrongdoing.
3:47
And a creative process does not become public property because somebody wants reassurance
3:51
about it. That's why you as a writer are
3:55
allowed to decide what your pro. What? Decide which parts of your process
3:59
are private, Maybe the whole thing. You're also allowed to discuss every
4:02
detail of it. Again, totally up to you. Writers have been talking
4:06
about their routines, their tools, rituals, research, false starts, editorial relationships,
4:11
strange working hours for centuries. Because, yeah, process can be fascinating.
4:15
I love hearing about people's process, but I'm not going to make them do it,
4:18
you know. So, yeah, I love it. Give me the weird stuff. Tell me about
4:22
your index cards, covered front and back with all the different colors of markers.
4:25
I love scene cards, by the way. But voluntary discussion
4:29
is different from being placed under suspicion and expected to produce evidence
4:33
of creative innocence. Substack describes the
4:37
issue as people pretending something written by AI was written by a person,
4:41
which begins with the assumption that the presence of AI can be identified
4:45
from the finished text and then interpreted somehow as deception.
4:49
But detecting a possible pattern in language does not reconstruct the
4:53
process that created it. The scanner does not know what you knew
4:57
before you started. It doesn't know whether you brought 20 years of professional experience
5:01
into the conversation, whether you supplied the original stories
5:05
and arguments, whether you already rejected 15 unusable
5:09
directions, whether you rewrote the entire piece, moved every section,
5:13
restored your natural language after the tool somehow managed to neutralize
5:17
it, or if you made every decision that affected what the reader
5:21
finally saw. They don't know any of that. It also doesn't know
5:25
whether you typed one vague sentence into an AI tool, copied the response, and published
5:28
it without reading it. Those are radically different things. Those are different processes.
5:33
A percentage applied to the finished text cannot tell
5:37
us which one of those things occurred. Pangram itself
5:41
acknowledges that false accusations of AI use can cause reputational
5:46
damage and other harm. And in the same breath, I'm sure its model
5:50
has a non zero error rate. The company says it works to reduce
5:54
those error, which is errors, which is appropriate. But accuracy is only part
5:58
of my objection here. Even a detector with a
6:02
very low error rate cannot tell you what authorship meant inside a
6:06
particular process. In other words, a machine cannot identify the author's
6:11
why, it can identify language patterns
6:14
that its model associates with different categories, but it cannot measure
6:18
how much the writer thought. It cannot measure critical
6:22
thinking, creative thinking, all the things that people are accusing AI of taking away from
6:26
us and outsourcing your thinking. Yeah, that's one of those phrases
6:30
people repeat as if we've all agreed on what that means, where the line is,
6:34
and how it can be measured from outside the person's own mind. Except you
6:38
know what? We haven't measured that. I think. I don't even think it's possible to
6:41
measure that. Suppose I use AI to organize a pile of notes
6:44
I created. Did I outsource my thinking? What if I ask it
6:48
to interview me about an idea until I clarify my own thoughts?
6:52
Is that outsourcing? What if I have already
6:56
written a full draft and then I use AI to identify repetitive sections?
7:00
What if I dictate an episode, receive a transcript,
7:04
and use a tool to clean up transcription errors? What if I
7:07
ask for 10 possible structures, reject nine of them, take one small piece from
7:11
the 10th, and then reorganize everything all over again? What if the
7:15
tool challenges an assumption I made and I decide the challenge is correct
7:19
and I change it? What if I ask for an argument to challenge my position
7:23
because I want to check my own thinking on the subject? Very, very meta.
7:27
Huge fan of that. The tool participated in each one of those things.
7:31
Yet the nature of that participation, the amount of human judgment involved,
7:35
and the effect it had on the writer's thinking, those are very different every
7:40
time. Only the writer knows whether the tool helped them think or
7:44
if it just gave them a convenient way to avoid thinking.
7:47
And sometimes the writer doesn't know right away. You know, we don't
7:51
know everything all the time. Which is the whole argument for self assessment. You may
7:55
have to look back at your recent work and admit that you have started approving
7:59
ideas from AI instead of developing your own. That's okay.
8:02
Self awareness Course corrections We're not perfect. It's fine.
8:06
You may notice that you're asking it for a point of view, maybe before you
8:10
even formed one, or accepting language because it sounds polished, even though without realizing
8:14
that maybe you don't sound that polished. I've now spent
8:18
12 episodes of the show talking about those risks. I'm not arguing that
8:22
every use of AI preserves authorship. Of course it doesn't. You can delegate
8:26
so much of the process that your role becomes little more than choosing among completed
8:31
outputs. And at that point, I think you should be honest with yourself about what
8:35
happened. But, you know, honesty with yourself,
8:38
that's the key part. The fact that someone else might have used
8:43
AI carelessly does not give a social media platform authority over your
8:47
process. It doesn't make you guilty until proven innocent.
8:50
And the fact that some people misrepresent AI generated work as
8:54
their own, that does not create a general obligation for every writer to
8:58
explain, reveal, or prove how they work. And in case
9:02
you haven't heard, panic about tools replacing human thought is not new
9:06
in history. In Plato's Bedros,
9:09
Socrates tells the story of an Egyptian king being presented with the
9:13
invention of writing. The inventor claims writing will improve memory
9:17
and wisdom. The king argues that it will produce forgetfulness because people
9:22
will rely on external marks instead of exercising their own memories.
9:26
And it will give them the appearance of wisdom without the real thing.
9:31
So writing itself was once the alarming
9:35
technology. Funny, right? The thing that would let people store thoughts
9:39
outside their own minds. God forbid, right? The concern. It wasn't
9:43
totally crazy. Writing did change how human beings remembered, learned,
9:47
communicated, and preserved knowledge. A tool can strengthen one ability
9:51
while reducing the need to practice another. This level of nuance is
9:55
missing from most of the current AI shouting match. We do
9:59
not have to choose between the tool has no effect on us,
10:02
and the tool destroys thinking. We can find
10:05
the middle ground. We can look at which parts of a process. It changes
10:09
what we gain, what we stop practicing, and whether we feel like it's
10:13
a fair exchange. Typewriters changed the physical
10:17
process of writing. Word processors made revision easier and allowed
10:21
writers to move, duplicate, remove, and restore text without retyping the
10:25
whole freaking page. Spellcheck took over part of proofreading.
10:29
Search engines changed research and voice recording changed
10:32
note taking and dictation. Writers responded differently because writers
10:37
work differently. Henry James dictated later work
10:41
to a typist. Mark Twain was an early major author. To submit a typewritten
10:45
manuscript to a publisher. I'd imagine that George R.R.
10:49
martin was talking talked about writing in his novels, in wordstar
10:53
on an old DOS computer, because it performs the function he wants without offering a
10:57
stream of automated corrections. Every one of those things, those are process
11:01
decisions made by the individual author. They're not reflections of Creative integrity.
11:06
No single tool provides rigorous or proves rigor,
11:09
just like no single tool proves laziness.
11:12
And using an older tool does not make the writing more human any more than
11:16
using a newer one automatically removes the human being.
11:20
The question is, what happened inside the collaboration between
11:24
the writer, the tool, and the work? My entire career as a
11:28
ghostwriter and book coach has involved entering someone else's creative process and helping
11:32
them turn their stories, expertise, ideas, and experiences all into
11:36
writing. Sometimes I interview someone extensively and create
11:40
a draft from those conversations, usually on a weekly basis.
11:43
Sometimes a client writes something first, and then I help them organize and develop
11:47
what's there. And sometimes I ask the right question that gets them past
11:51
the right answer that they thought they were supposed to give, and I get access
11:55
to the real answer that's at the core of it all. That type
11:59
of interview actually is my favorite. When they think one thing and then I tell
12:01
them, hey, look, we found something else. Yay. Gold mining. Human assistance
12:06
has been part of professional writing for a very long time. Editors suggest
12:11
language researchers collect material. Interviewers draw
12:15
out ideas and magic. Writing partners challenge
12:19
each other with scenes, scene ideas. Coaches identify missing pieces.
12:23
Typists and transcriptionists convert speech into text.
12:26
And ghostwriters, when the process is done ethically
12:30
anyway, they help people communicate work rooted in the person's own experience,
12:34
expertise, and point of view. Those are all examples
12:38
of tools and processes. Those relationships can also go wrong.
12:41
Of course they can. They're human. A collaborator can overpower the writer. An editor
12:46
can remove the parts that made the work distinctive. Remember the I think it
12:50
was the Emily Dickinson episode we did a few weeks ago? Here, a ghostwriter
12:54
can impose a voice that doesn't belong to the authority.
12:57
The presence of a human doesn't settle the ethical question. Again,
13:01
you have to understand your process. You need to know which parts
13:05
require your direct involvement because they contain the reason you're writing.
13:08
The decisions that define your position are included in that. The stories
13:12
only you can tell and the language through which readers recognize you.
13:17
You also need to know which parts can receive support without compromising the
13:21
whole thing. Your list is going to different be different than mine. It may different
13:25
be different from one project to the next. For one piece, maybe you
13:29
use AI for research, organization and nothing else. For another. Maybe you
13:33
use it as a developmental interviewer, like your own ghostwriter.
13:36
You may decide that generating prose crosses your line, while another
13:40
experienced writer sees the generation of writing as raw material that
13:44
they can reshape through a lengthy editorial process. You do not
13:48
have to Adopt other people's boundaries and they do not have to adopt yours.
13:53
Now what concerns me is this attempt to replace this personal understanding.
13:57
I mean, look at all the examples I just gave. Look at how specific and
14:00
personal each of those examples and those historical stories were.
14:03
Now we're going to try to replace that with a platform
14:07
approved process and then we're going to attach to it a public score
14:11
that encourages strangers to make judgments that they don't even have like 1% enough
14:15
information to make. Freedom of speech is not the only creative
14:20
freedom in our domains domain of interest. As writers, as artists, having a
14:24
free path to arrive at that speech is equally important. Freedom of
14:28
creative process begins with personal understanding and governance over your own
14:33
work. So before you defend your process to somebody else,
14:37
make sure you can explain it to yourself. Where do your ideas begin?
14:40
Which parts of writing help you discover what you think?
14:44
What work are you willing to delegate? Which decisions need to
14:48
remain yours under all circumstances? Where has AI actually
14:52
strengthened your process? Where has it started to replace a
14:56
skill or habit or judgment that you still want to hold on to?
14:59
And what, if anything, do you choose to tell readers about any
15:03
of this? Choose. That's the important word.
15:06
Your process. It might be interesting to other people, it might be useful to other
15:10
writers, it might become a part of how you teach or write or connect with
15:13
the audience. But it's yours before it's
15:17
anyone else's information. So let's
15:21
give some homework, as we always do this week. Create a private process
15:24
map for one piece of writing. Start with the first idea and trace
15:29
what happens until publication. Write down every tool you use. Tool,
15:32
person, conversation, research, step, draft, AI,
15:36
interaction, editing, rejecting everything
15:40
that goes into creating a writing piece tool. Just kind of jot it all
15:43
down and then answer these three things.
15:47
Where did I do my most important thinking?
15:50
Where did a tool or another person genuinely help me?
15:55
And where did I give away a decision I should have made?
15:59
Don't create this map to disclose it if you don't want to. You don't have
16:03
to put it on substack or social media. Create it so that you know your
16:07
process well enough that the cynics and the censors cannot define it
16:11
for you. Know thyself, know thy writing,
16:15
know thy process, and decide for yourself what remains yours.
16:20
All right, that's it for today's, I guess, kind of,
16:24
what do they call it? Soapbox. That's the word I was looking for. Soapbox.
16:28
Speech on freedom of creative process. Hope you enjoyed that. That's it for
16:32
today's. Human Print. Until next time, Use the tool if it
16:36
helps. Just don't hand it the pen.
16:45
And that's it for this episode of Human Print. If you want a practical
16:50
framework for protecting your voice amid the AI Noise, please follow the show.
16:53
And before you go, take a minute and ask yourself, as the
16:57
AI noise grows louder and louder, which areas of yourself
17:01
do you need to protect the most? I'll see you next time,