Subscribe
Share
Share
Embed
What if you could save 12+ hours per week in your business? Being a solopreneur sometimes focuses too much on the “solo” part: doing all the jobs, figuring things out yourself, and spending too much time in your business. But we didn’t start out own solo business to spend all of our time at our desk.
We did it because we want freedom: to travel; to spend time with our family; to watch a movie in the middle of a week day. That’s why Streamlined Solopreneur exists.
Each week, host Joe Casabona talks about how you can build a better business through smarter systems and automated processes. He does this by bringing on expert guests, and sharing his own experience from years as a busy solopreneur parent — so that being a solopreneur feels…less solo.
With every episode, you'll get insights, great stories, and 1-3 actions you can take today to improve your business processes and spend your time the way you want.
Joe: In early November,
I ruffled some feathers.
I made a couple of posts about.
AI and how it should
and should not be used.
The first one was me saying
that telling AI to act like a
doctor makes it as much a doctor.
As selling George Clooney.
To act like a doctor.
But the second post.
The second post.
Really got under a lot of people's skin.
Where I said, using AI to write your book.
Is like using a car.
To run a marathon.
Now I'll admit that maybe
that's not a perfect analogy.
But one of the people who I got
in the spirited debate with was
my good friend, Alistair McDermott
over at the recognized authority.
And he suggested we act like
true podcasters in gentlemen.
And get on an episode.
And have a debate.
And so we recorded it on his
channel over on his live stream.
And he provided me with the audio.
So in this very special bonus episode
of how I built it, I'm going to give
you the debate that Allister and
I had, I think it's really good.
Uh, I don't want to spoil anything,
but there are points where we
fully agree and points where we
agree to disagree as we continue.
To hash out and fully bake our
thoughts on this very new landscape.
For business owners and creators.
So I hope you enjoy this episode.
This bonus episode, this debate.
Between me and Allister McDermott.
Let's get into the intro.
And then the debate.
Hey everybody, and welcome to How I
Built It, the podcast where you get free
coaching calls from successful creators.
Each week you get actionable advice
on how you can build a better creator
business to increase revenue and establish
yourself as an authority in your field.
I'm your host, Joe Casabona.
Now let's get to it.
Alastair: Hello and welcome
to the recognized authority.
I'm Alistair McDermott and
my guest is Joe Casabona.
Today, we are going to discuss can
and should we use AI to write for us?
Welcome, Joe.
Joe: Thanks for having me, Alistair.
I'm excited to
Alastair: talk about this.
Joe, you are somebody who's
been on the show before you were
on, I think it was episode 36.
Yeah.
Episode 36 back in October of 2021.
So two years later, uh,
we're here again chatting.
And so you have ruffled some feathers
with what you've been saying about AI
on LinkedIn and people are blocking
you and stuff like that because you're,
uh, you're really poking poking folks.
So can you tell me a little bit about what
you said about AI and let's get into it.
Joe: Yeah.
So this was, so the impetus for this chat
was actually, uh, the second apparently
inflammatory post I've made about AI.
Uh, the first one.
Was, um, telling AI to act like a doctor
with 20 years experience makes it a
doctor as much as telling George Clooney
to act like a doctor makes him a doctor.
People got real mad at that.
Um, and so I decided to follow it up with,
I saw a post somewhere in some community.
about how this person was writing
25, 000 words a month using AI.
And I'm like, well, that means you're
not writing 25, 000 words a month.
AI is writing some of those words.
And so I posted across social media saying
I used AI to write my book or I use AI to
write a blog post a day is like saying.
I used my car to run a marathon, and a
lot of people, uh, got mad that you and
I had a spirited debate on that post.
Alastair: We certainly did.
I think that the analogy is not perfect,
like all analogies, but there is some
truth to what you're saying as well.
So that's why I wanted to get you on
the show and to just chat about it,
rather than just blocking you, you know?
Joe: Yeah, absolutely.
Right.
Reason.
I'm all about reasonable discourse.
And so when someone is being
unreasonable, I, I no longer go for it.
So the, the, in, in one case, this person
just liked to hear themselves talk.
And so I would say something and
then they would say something and
then I would respond and then they
would respond with a non sequitur.
And I'm like, well, you're
not worth arguing with.
Because you are not,
we're not arguing, right?
Or we're not debating,
whatever you want to say.
Uh, but yeah, you and I
point counterpoint, right?
Reasonable discourse.
Alastair: Well, of course, as
podcasters, you know, we, we do
like to talk a little bit as well.
Joe: Yes.
I won't, I won't deny that.
I like hearing myself talk, but usually
I like hearing myself talk about
whatever is relevant in the conversation.
Alastair: So let's talk about this then.
Um, all right.
So let's talk about the first one, right?
I'm using AI to write my book.
Or I'm using AI to
write a blog post a day.
Yeah.
And you're saying that's like
I'm using my car to write a
marathon, to run a marathon.
But, so the issue for me is, well, you
can be using AI to help write something
and it not be like getting in your
car at the start of a marathon and
just driving to the end line, right?
You can actually use it in a way.
And so when you dig into the nuance,
you, you, you really do need to dig in to
figure out how they're actually using it
because it's possible to write something.
I mean, you can't like, you can
just give chat GPT a bunch of
prompts and have it spit back out.
An extremely mediocre, almost by
definition, mediocre, because it's kind
of the average of all of the information
that it's learned is what it's outputting.
So, but yeah, and I think that most
people who are using it to create content.
I would like to say aren't
using it that way or definitely
shouldn't use it that way,
Joe: right?
Yeah, so this is where I
think we're in agreement.
And yes, my so this makes
my post ambiguous, right?
Because, um, I envision,
I've like the example I told
you where someone was right.
I'm writing 25, 000 words a day, a
week or a month, whatever with with AI.
I've also talked to people who are
like, Oh, yeah, I've used AI to
write full chapters of my book.
And I'm like, you didn't write the book
then, like you didn't write that chapter.
If those are not your
words, you didn't write it.
And so like, if you use your car to
drive to where the marathon is happening,
and then you run the 26 miles, no
argue, no bones about that, right?
Absolutely.
Definitely use your car.
Uh, unless you're like a maniac
who like runs to the marathon
and then runs the marathon.
I know people like that, but let's say
most of us are normal human beings.
You're going to drive yourself
to the starting line, right?
Just like use AI for research,
ideation, outlining, maybe, maybe
giving you a jump off point.
Those are fine, right?
Because then you're going
to take that raw material.
And turn it into your words and your
book and I don't have a problem.
I do that.
I do that.
So I don't have a
problem with that per se,
Alastair: right?
So the issue for me is if, if, if
somebody is saying that they're
refusing to use AI to help with
the process of content creation.
I think that as a content creator, that
is like an accountant saying, Oh, I
don't use spreadsheets or calculators.
You know, I think you're absolutely
nuts not to use AI, uh, because of how
much that it can help and support us.
And the way that I feel about it
is it can make somebody who is
good at producing good content.
It can make them much more
effective and efficient.
at creating good content and maybe
even raise the quality of that
good content up a notch because
it's able to do things much faster.
Like some of the stuff that some of
us might see as BS, like spelling
and grammar and things like
that, they need to be done right.
But, you know, we don't like to be, you
know, going around fixing that kind of
stuff when we could be writing a new
chapter or coming up with new ideas.
So that's where it is for me
is like, I think you're nuts if
you're not using it for that.
Joe: Yeah, agree wholeheartedly, right?
I agree wholeheartedly
with that statement.
In fact, um, I have a course
on LinkedIn Learning called
Generative AI for Podcasters.
And I talk about everything that
you can use generative AI for.
One of those things is creating
a listener avatar for your show.
And let me tell you, I did this for my
show to see what it would come up with.
And if I didn't know AI generated
it, I would have thought they
were describing a real person.
Uh, and so that gives me clarity
for who I'm talking to, especially
in the beginning, right?
If I don't have an audience,
I need somebody to talk to
until I build that audience.
And so, I agree wholeheartedly
with that statement.
I used AI this morning
to write a press release.
'cause I don't think I can write
press releases and they, I don't know,
maybe they're just more factual than
blog posts and then like, I'm having
a friend punch it up for me to make
it not sound like ai, but I don't, I
wouldn't say I wrote that press release.
Right.
I would say I used AI to
write a press release.
Alastair: You see, now we're
getting into semantics and Yeah.
The, the , that's, that's, I I, I hate
when these arguments get into semantic.
Now, I think when I see AI creating
stuff like, and I'm going to
read something that AI wrote.
So are you ready to unlock the clandestine
power of AI and delve into a realm of
endless possibilities in the modern
digital landscape, staying ahead as
the new normal, unlock the secrets
of, I think that that is such BS.
That is just crap content.
And I hope you didn't get
any of that in your shoes.
I mean, you can use AI to write good
content that doesn't sound like that.
And, and that's where, like,
that's where I, I, I worry.
Cause I see a lot of people using
it and creating rubbish content.
And then I see people saying things
that sound like, don't use AI.
And that's, that's my concern
is like, well, yeah, do use it,
but use it in the right way.
Joe: Use it judiciously.
I think, I think we both,
and we kind of came to this
conclusion ahead of time, right?
I think we both are closer.
In our beliefs than we
initially thought, right?
Uh, there was another person who
commented saying, like, If my audience
can't tell the difference between
what I write and what AI wrote,
Why am I going to waste my time?
And I, very bluntly said, That's a
problem with you, not your audience.
Like, if you can't, If you can't
write in your own voice, Or a
voice that's differentiated from
whatever you just read there about
clandestine something somethings.
Right?
That means that you're a bad writer.
And if I'm a bad runner, and I
want to get better at running,
I'm not going to drive everywhere.
I'm going to run more.
If I don't care about writing, fine.
But I'm also not going
to call myself a writer.
Alastair: I think that one thing that
is super important to me is using
AI to get faster and more efficient.
And be more productive.
And I, I, a friend of mine challenged
me to go and quantify, because
there's, there's a great book.
I think it's called how to measure
anything, but he said, can you
quantify how much AI is, how much more
effective it's making a more efficient.
And it was making me about 40
percent more efficient at writing
content, creating content.
See, now, now we got to get specific
about the words that we use here.
Yeah.
Um, but I'm still writing that content
and the ideas are still coming from me.
Even though it's applying a lot of the
grunt work, and this is where, you know,
um, like, I, I think of, like, if this
was okay, I just made some, I just made
some lamb kebabs in the kitchen, right?
I was talking to you.
I was down.
I went downstairs cooking
and I left the marinating.
So when I was, when I
went down to cook those.
I chopped some onions in a, uh, a
spinning, uh, chopping, uh, device.
I don't know what you'd call it, but
it basically shredded the onions down
super quick in like five seconds.
Boom.
Done.
I wouldn't have made it back in time
for this call if I had done it manually.
So, uh, so is that to say
that I didn't cook that meal?
You know?
So no, it's just, I used a tool
that made me more effective.
That's, that's the point.
Like, I still decided what I
was going to make, you know?
Joe: Yeah, but, so, but I think,
right, where we are, we are, we are
different, differing in our analogy
is, my analogy was more like you going
to a restaurant, buying lamb kebabs.
And then saying, I made these, right?
Yeah.
Alastair: Okay.
Right.
That's my road.
Yeah.
The other issue that comes into this, when
it comes into like writing and creating
content is there is a learning process
that occurs in your brain when you write.
Because you have to formulate the
thoughts and put those thoughts together.
And this is where I could see it
potentially being dangerous for people who
are not experts in the field, because if
they're using AI to generate and to write
for them, they're not going through that
learning process where they're actually
putting the ideas together in their head.
And, um, I think another friend
of mine said something like.
Editing is learning or editing
is thinking, I can't remember
something, something like that.
It could be, could be Jonathan Stark
or something like that, but the idea
that, so the idea is, you know, if you
already have great ideas and you've
thought a lot about your topic and you've
maybe written about it, then using AI
can really speed up what you're doing.
But if you're not at that
stage, then it may be.
Potentially going to damage the journey
that's that's a real concern for me for
somebody who's trying to become an expert
on an authority in their field, which
is something that I think a lot about.
How do you think about that?
Joe: Yeah, I agree wholeheartedly, right?
And like I have.
So I have three small kids, right?
There are 63 and two,
um, my three year old.
Can do, can do thing, more things at
three than my oldest could do at three.
And it's not about aptitude
and it's not about skill.
It's that my wife and I did
more for our oldest because we
didn't think she could do it.
So we just did it for her
instead of telling her to do
it herself or try it herself.
Whereas, we learned, and with our son, he
tries to, he tries to do something first.
We make him try to do it first.
And so he developed the skill faster.
So when you're talking about,
you know, using AI to do things.
Instead of doing it yourself,
that's exactly what I thought of.
If I'm tying my shoes for my
daughter, she's never going
to learn how to tie her shoes.
Alastair: So then the question becomes,
well, what if we have a society, and we'll
take that analogy, where we have machines
that will just tie our shoes for us?
And it's super simple and very quick.
Do we ever need to learn
how to tie our shoes?
Joe: Yeah, that's a good question.
Right.
Or maybe we just all wear Crocs
like in the movie Idiocracy.
Right.
Um, which Crocs was not known
when that movie was made.
I don't know if you know this.
Here's a quick tangent.
Uh, Uh, the wardrobe designer for
Idiocracy was looking for just kind
of generic futuristic looking shoes
and they found this little known
company, um, and it was Crocs.
And they're like, these, nobody will
wear these ever, like they're so hideous.
And now they're probably
the most common shoe.
Um, everybody in my house has
Alastair: them.
Yeah, the product
Joe: placement that did it.
No, it wasn't, it like wasn't
even product placement.
She just sourced the shoes for the movie.
Crazy, right?
Um, so I mean maybe the movie helped
it get big, but it was not known, and
the wardrobe designer didn't think.
Anybody would be wearing them.
That's why they went with those shoes.
Um, point being, right.
Uh, my father in law thinks a
lot about that with maps, right?
He like knows how to read
a real map and I don't.
Um, and you know, he's like,
if an EMP ever goes off.
Most of the world is screwed.
Unlike me, wholeheartedly.
Like, if computers didn't
exist, I'd be dead.
Basically.
Um, And so, I think a
lot about that, right?
Like, are we moving towards some
futuristic, like, WALL E esque world
where we're all in, like, these movable
chairs and we're all fat and just watching
TV because we don't need to do anything
because machines do it all for us.
Like, do we lose humanity at that point?
Like, do we lose our
humanity at that point?
Alastair: Yeah.
And, uh, you know, I mean, maybe
that's just beyond the scope
of what we're talking about
because it's just, that's like,
Joe: that's so far.
Yeah.
That's a real heavy, to get that heavy.
Alastair: Um, but I mean, like, I'm sure
that people said the same thing when
the calculator was created, you know?
Oh, people aren't going to do
maths in their head anymore.
And that's, you know, that's.
Uh, I even heard somebody talking about,
I think it was Erasmus, the philosopher,
uh, and mathematician, uh, complained
when the printing press was invented
that there was going to be too much
content created and that the quality
of content was go, was going to, uh, be
mediocre and that, that it would be harder
for good quality content to surface.
And that was, you know, when was that?
The 1600s.
So, um, I'm showing my, my
terrible history here with
not knowing the date of that.
No, that's
accurate
Joe: though.
Uh, I'm sorry.
No, it was the 1400s.
The printing press was invented like the
same year Leonardo da Vinci was born.
And I only know that cause I just
started reading Walter Isaacson's
book on Leonardo da Vinci.
Alastair: Yes.
Right.
And he's a fascinating character.
And, and, uh, I've been in his
house that he had in, um, in France.
Uh, he, he lived beside the French
King in, um, in the lower valley.
Oh yeah.
Fascinating place to visit.
Um, but now that's a tangent.
So the, so whenever this new technology
has come along, people, people are saying,
okay, the, you know, the, the end is nigh,
but let's talk about the practicalities.
Yes.
The, the situation where somebody
wants to use AI to write when,
when is it a good idea to use it?
When is it not a good idea to use it?
Joe: Yeah, I think, I think like
with any technology we are learning.
We're learning, right?
It's a new technology.
And so, uh, just like when phones
first came out, it didn't have all
of the screen time, focus, filters,
parental controls that we have now.
So we're definitely in like
the, call it the AI renaissance.
Maybe we're all trying to, maybe
it's pre Renaissance, right?
Cause Renaissance is like
the really good age, right?
But so we're like pre Renaissance,
we're just trying stuff now.
Um, and so I'm taking a very hard line
skepticism of people who say, Oh yeah,
you can literally use AI for anything.
Because I think that what we
can't use AI for is, is anything
that requires the human element.
telling stories, writing, right?
If I just want, here's an example, right?
I tried this, right?
Cause I'm not just like
someone who says, don't do it.
And I'm not going to try it.
I'm going to try it.
So I wrote a primer for AI.
So I prompted.
Chat GPT for, and I said, write
me a 700 word blog post, basically
explaining AI and large language models.
And like, obviously it
got the facts right.
I knew the facts already because
I had done my own research.
Um, I had done my own
reading on the topic.
Uh, but it was just like a
flat telling of what AI is.
And so I had to punch it up.
I talked about Star Wars and I
added my own personal anecdotes.
And what I kept from that blog post.
Was basically like two bulleted
lists, uh, of like what large language
models are and, and some ideas that
you, that you could use for them.
Did I, did I write that blog post?
I wrote, I don't know, 90 percent of it.
Um, and so would people read
it if I just took what ChatGPT
gave me and, and published it?
I don't think so, because there's no hook.
There's nothing that grabbed me.
Alastair: So what if you
asked it to add a hook?
Joe: It still wouldn't be my hook, right?
Like it would be like,
Alastair: Then you can say, you
know, I don't like that hook.
Give me three other options.
What about using the hook that does
something else and suggest one?
Joe: It's still not me writing it.
Right.
It's still not my voice.
Right.
And so
Alastair: you see that that's.
That's the question.
I mean, you can, like the way that I use
it, I use it in a very iterative way.
So I'll get feet, I'll, I'll
get output and I'll say, um, I
don't like that call to action.
I don't like that opening hook.
I suggest me another hook
or use this as a hook.
And, and am I not then writing
Joe: it?
No, I don't think you are.
You're telling someone else to write it.
The chat GPT is basically
your ghost writer.
Alastair: Yeah.
So, I mean, that's the question.
And like, I think they're like,
there becomes, there becomes a line.
And I think that this is a spectrum where.
If you iterate on something enough
times and ask for enough changes,
that it becomes your words.
I mean, it has to,
Joe: but they're not your words,
Alastair: right?
Well, I don't know about that.
I mean, it's someone else's
Joe: words.
That you extracted from
the large language model.
Alastair: Yeah, so, I mean, the, I mean,
I guess this is where we disagree, because
I think that at some point, like, it's
a question of what are you feeding in?
Like what are the instructions and how
detailed are those instructions you're
feeding in, in terms of asking you
to generate something, because some
people will feed in a one line prompt,
some people will feed in a 10 line
prompt, some people will will feed in
a five line prompt, and you know, 3000
words that they've already written.
Generated somehow, for example, both of
us, uh, probably generate a lot of words
through, uh, through podcast transcripts.
So we could take the transcript of this
conversation and I could say, just pull
out Alistair's side of this conversation.
Now go through Alistair's, uh, what
Alistair said and summarize that.
Give me a detailed summary of that.
Now write that as a blog post, right?
Are you saying that's, that's
me writing that or not writing
Joe: that?
Okay.
No, I wouldn't say you're writing it.
I would say it's your
words being used, right?
Alastair: You see, that to me
sounds like semantics because...
Joe: But it's the difference between like
a biography and an autobiography, right?
Alastair: Um, yeah, I guess
maybe we're not going to agree
on this because I think that...
Having a biography, I guess it's
kind of like maybe it's an authorized
biography where the, where the person
is standing over the shoulder of the
writer and saying, Hey, yes, I do.
Or I don't, you know?
Right.
Joe: Yeah.
I, yeah.
Right.
Isaacson interviewed, uh, he didn't
interview Steve jobs early in our da
Vinci, obviously, but he interviewed a
bunch of people who knew him and then has
footage and all this and put something
together for the Steve jobs book.
We'll say.
I wouldn't, even though a lot
of those words are probably
Steve Jobs and it's his life, I
wouldn't say Steve Jobs wrote that.
Yeah.
Alastair: So, so, okay.
Apart from the fact that we, we maybe,
uh, are in disagreement over that,
where I think it's important is, is like
right now, this is not that important.
I think because The words that it's
generating without any editing are
just not good enough to pass for human.
Yeah, we, we, we wholly agree on that.
Yes.
But at some point they will, at some point
it's going to get good enough for us to,
to not be able to tell, or for, you know,
99 percent of us not to be able to tell.
Right.
Joe: I think you're
probably right about that.
Alastair: And then we have other things
that come along like custom GPTs.
I did a live stream about this
last week where I made a custom
uploaded four of my books to it.
So I now have a custom chat GPT bot
that knows what I've said about a
topic and knows my writing style.
So if I then say, you know, um, give me a,
you know, give me a blog post about this
topic based on what I wrote in that book.
It's able to create a new blog post that
kind of synthesizes the ideas from that.
And it's, and it's going to be in a
fairly close, it's still not going
to be perfect, but it's going to
be fairly close to what I said.
Joe: It's going to be, yeah, it's going
to be fairly close to what you said, but
it's not going to be what, it might not
be what Alistair in 2023 would write.
Because since you wrote those
books, I know one, you just wrote
around this time last year, but
those previous books will say.
You were a different person, right?
Some people say that we
change every two years, right?
Like yeah, you look back at
yourself two years ago, and you
feel like you're a different person.
So That's where AI and chat
GPT will always fail, right?
It could have all the words I've ever
written But when I'm 40 years old,
two years from now, I'm gonna have an
A preteen and two and two other small
children and I might be living in a
different house and I'll have a bunch of
different life experiences that I will
use to relate my ideas to other people.
And so I think I wanna, I wanna say this
thought before I lose it, because when
you were talking about prompting and
iterating, it's the, I think that's the
difference between like being a visionary.
And being the actual worker, right?
Steve again.
Well, I'll just use Steve Jobs again.
I guess Steve Jobs was a visionary for
the iPhone, but no one would ever say
he physically made the iPhone, right?
Because he, he got it.
He tested it.
A bunch of people provided feedback
and the iPhone that came out was
from innovations from engineers.
And it was really the first
thing of its kind to take
this example one step further.
You saw a lot of copycats
after that, right?
That weren't as good as
the iPhone or whatever.
Pick your favorite phone, I guess.
Um, weren't as good as the iPhone,
uh, but were imitations based on
what they knew about the iPhone.
So I, I, I'm.
Maybe pulling together an analogy
here, um, that's not, that's kind of
half baked, but, you know, I think
it's kind of the difference between,
like, the visionary doing the work, and
then, like, copying the original work.
Alastair: Yeah, so, I guess, I'm, like,
I'm trying to, I'm still trying to figure
this out, because, for example, I, when
I drive, I sometimes narrate and record
or dictate, and quite often I will do
that into the otter app on my phone.
Joe: So I use audio pen.
Yeah.
Alastair: Yeah.
And so what I'll do is I'll, I'll record
some stuff in there and I'll like, that
might be an idea for a blog post or it
could be for a podcast episode or just
something about my to do for, for the day.
But quite often I'll take that text
and run it through chat GPT and
I'll say, clean this up, uh, give
it to me in bullet points or give
it to me in the form of a table.
Um, you know, I asked her for that today.
I was, I was looking at updating a
product and I said, I want, I want
you to give me the, you know, the
new pricing and the new, uh, product
features in the, in the form of a table.
So that, that kind of thing, like
where I'm dictating in, like I
feel, well, that's me writing.
You know, um, even if it does some
significant cleanup, those ideas are still
coming from me and not only that, but when
it's doing the cleanup, the cleanup that
it's doing is based on my suggestions.
It's not like it's coming up with,
if I say, you know what I want you
to, uh, you know, that's too wordy.
Like the one I like, I always say,
make it more concise, more directly
written and remove filler and fluff.
Right.
And when it does that, lo and behold, it
starts to sound a lot more like I write.
And I don't know if that's because of
like, I write more, more tersely because
of an engineering background or whatever,
but I tend not to write in, in, in that
kind of way where it's quite wordy.
So, but, but I feel like I am writing
when I, when I, when I give it those
instructions, because what's coming out
is my ideas in my voice, sounding like me.
Uh, and, and has been edited by me,
maybe through an, an iterative editing
process where I talk to the AI.
And by the way, quite literally, I mean,
uh, I use chat GPT on my phone where I
use the voice dictation input feature.
So quite often I'm actually
having a conversation literally
where I'm speaking to it.
Uh, and, and yet I feel like I'm writing
that and, and I think that maybe this
is a semantics thing, but, uh, you
know, do you, do you disagree with me?
Joe: I, well, they're
definitely your ideas.
Right.
I would, when I do this with
audio patent, right, and maybe
I'm not using the right tool.
Right.
Um, I never feel like what it
comes up with is suitably me.
Right.
It takes what I say, and it
turns it into some language.
And if it's just like a social
post, like, maybe I don't care.
Maybe I'll just throw that up.
But I have very rarely felt like what
I got back from an AI Was something
that I felt I couldn't have done a
better job writing or editing, right?
And so, um, this is probably,
this is obviously very subjective.
Um, I think that, I think that if
you are dictating the words, Right?
This could be semantics, right?
If you listen to a bunch of books on
Audible, are you reading those books?
Right?
Or like you're cheating
a little bit, right?
Because like you can listen to
it at 1x and you're hearing it
faster than, than you're reading
it and maybe, but maybe you're not
comprehending it as well either, right?
So, um, Like, I think that we
need to consider what we're
losing with the shortcut.
And so if you're dictating and
iterating and you eventually get
something that's good, what have
you given up to get to that point?
I think that's probably,
and that's gonna be
Alastair: different for everybody.
So, right, to take the argument
that you made an argument, which
is effectively The output that
I've seen so far isn't good enough.
Now, my counterpoint to that is,
that's irrelevant because at some
point soon it will be good enough.
Whether that's chat GPT 5 or GPT 4.
1 or whatever.
At some point that output is probably
going to be good enough for, you know,
99 percent of Of uses now, I think
that, you know, there's always going
to be that last, the last 1 percent
that it's, it's never going to going
to, um, that it's never going to get
to in terms of quality of output.
Now, what I'm talking about here,
by the way, is in writing style.
Uh, I think that in terms of actually
formulating and having new ideas, And
synthesizing new ideas from nothing.
I think that that's like,
that's not generative AI.
That's not, you know, chat GPT and
those models, there might be some
other type of, of AI that comes along
and can do that, but, but that's not
really what we're talking about here.
Right.
Joe: Right.
Yeah.
And in this moment, right, the thing I'm
running with now is, is you talking at
an AI and coming up with some writing
and then you iterating over that.
Right.
I think you're probably right.
Right.
Especially with these GP, like
these custom GPTs, I could feed.
One, I've been writing on the
internet for 20 years, right?
Now, the way I wrote in
20, in 2003, was very bad.
Uh, so maybe I give it the
last five years of me writing.
It's probably gonna do a pretty good job.
A pretty decent facsimile.
Of how I would write something right
and I still wouldn't say I wrote
it because I didn't write it right.
It's I covered the 26
miles, but I didn't run it.
So I shouldn't be celebrated for it.
I think that's the big thing,
Alastair: right?
Yeah.
And so for me, if I, if
I dictate for an hour.
As I'm driving on a topic, right?
And let's say I dictate an hour is about,
um, like 000 words, something like that.
Maybe, maybe not, not that many, but
in a somewhere around that, right.
And if I dictate for an hour about
a topic and I, and I come up,
but let's say it's 8, 000 words.
Yeah.
And then I put that through generative
AI, it mightn't go through chat,
GPT might be too much for that.
Um, but I boil that down and I clean that
up and I use it to reformat it and put in
the nice headings and things like that.
My argument is I wrote that.
Those ideas came from me.
The, the, the order that things
came in came from me, the ideas.
And the fact that I had an editor
come around afterwards, humans have
been doing that forever, where we've
been giving content to editors and
getting them to clean up after us.
We're still saying we wrote it.
This
Joe: is the most, this is your
most compelling argument, right?
Because you've put in
an hour's worth of work.
And like, yeah, my book,
my last book, right.
Uh, uh, HTML and CSS, a
visual quick start guide.
I wrote, I say, I wrote that book.
I'm the only author on that book, but my
editor reworded large swaths of that book.
And this was actually
something I wrestled with.
I was like, can I even say I wrote this?
Like he rewrote this basically.
Can I say I wrote it?
Um, but ultimately we both
like, that's the job he's in.
Right.
We both agreed.
That I wrote it because I put together the
outline, I ordered everything, I wrote the
code, I put the ideas down, and then he
restructured it in the format of that book
Alastair: series.
And the other argument there is, if
you hadn't done your part, then your
editor could not have done their part.
You're right,
Joe: he would not have
been able to do his part.
And,
Alastair: and, and, and so the issue
right now is, is I think that what we
are creating with AI for the most part,
if you're an expert in your field,
what you are going to create with AI
is going to be vastly different than
somebody who is not an expert in the
field because they're not going to know
when it gives you some sort of mediocre,
bland output that that isn't good enough.
Right.
Whereas an expert will look at the
output and say, actually, it's missing
something really fundamental or something
that I feel is fundamental to my
distinct point of view on this topic.
And once you add that, then I think
you've got something valuable.
And then it, it, it, it then goes to
the kind of like the value pricing
versus hourly pricing discussion where,
um, and, and just to summarize that
really quickly, the, um, the, that
argument is you shouldn't bill people.
Hourly because it shouldn't
matter how long something takes.
It should matter how valuable
the output is to to your client.
And in fact, like, for example, if I'm
editing a video for somebody, let's say
it takes me two hours and I'm able to find
a tool that lets me do it in one hour.
Should I continue to do it the slow
way and charge them twice as much?
And that's where the ethics
of hourly billing come in.
And that's why Jonathan Stark talks
about that on his show, Ditching Hourly.
Um, but so, so in the same way,
like, You know, should we avoid
using these tools that help us
to do these things much quicker?
I mean, and, and, um, I think that there's
a connection between those concepts.
So I haven't articulated that very
well, but I think that, you know, that
the like, we shouldn't necessarily
have to do things the slow way in
order to claim the value of the work.
It should be about the
value of the output.
And is this good?
I put not.
Did you write this or did you
run the marathon manually?
Like if it depends on the goal, right?
If the goal is to run the
marathon in order for to have the
achievement of the health and the
fitness, then that's one thing.
If the goal is to get 26.
12 miles, yeah, 26.
1.
Yeah.
If the goal is to just travel a, to be 26.
1 miles, then yeah, if we can, you know,
if we, if we can use a jet pack to get
there, um, we, we've got there in the end.
So that's all that matters.
You know, that's, that, that's
my argument on that is like,
it's what's the value of the
Joe: 26.
Alastair: 2.
Okay.
I should, I should know this.
I volunteered at Dublin
Marathon two weeks ago.
Um, anyway, uh, So, so yeah, so, so for
me, it's, it's like, what's the value
of the output and, and is the output.
And then I guess the argument is
who's, who are we talking about the
value to, are we talking about the
value to the consumer of the, of the
output or to the creator of the output?
Because then it brings back in the issue
of learning from the writing process.
Right.
And that wraps back around to
what we talked about at the
start learning from writing.
Joe: Yes.
And I, I, I think you're right.
Right.
Again, like if, if your goal is just
to cover the distance, do whatever
you need to do to cover the distance.
Right.
But I don't think just
covering the distance.
Should be celebrated.
I think this is really
where I'm at, right?
If someone's gonna wear the badge of
honor for doing a marathon, for running
a marathon, then they need to run it.
Just like if somebody is wearing the
badge of, I authored this book, right?
Then they need to Let's take
your dictated example, right?
I will, I will cede this point to you.
They need to get all of
the ideas down there.
The book needs to be fully their ideas.
They can't say chat GPT.
I'm writing a book about podcasting.
Chapter one is how to start a podcast.
Right that chapter and
then slap their name on it.
This is I am I am
Unmoving in that opinion.
Alastair: Okay.
So what about they say chapter one
is going to be you know How to write
a pot how to create a podcast here
are The five sections I'm going to
have in this chapter, can you think
of anything I might have missed?
Joe: That's ideation and I think
it's, I think chat, GBT is,
or AI in general is great for
Alastair: that.
Mm-Hmm.
Yeah.
And, and so what we're talking
about then, like, see there's
a lot of nuance to this really.
There's a lot there.
Joe: This argument there is
just like, I, I mean like
nothing's black and white, right?
Mm-Hmm.
like, um, but there is, right?
Am I, am I cheating by
wearing running shoes?
In a marathon?
No.
It'd be crazy to run barefoot.
Alastair: Do you know that there
are running shoes that have these
kind of composed blades of carbon
that actually make you run faster?
That costs lots and lots of money,
but that Oh, no, I didn't know that.
Yeah, they've helped them break
the two hour mark on the marathon.
Yeah.
So, um, so there you go.
I don't know.
I don't know how long a marathon is,
but I know that you can get shoes that,
that will actually make you faster.
Right.
Yeah.
Joe: Well, it's just like,
you know, in baseball, right.
Is, is taking performance
enhancing drugs, cheating.
Like can you, can you say you hit
those, you still have the mechanics.
I don't know.
Yeah.
Alastair: Yeah.
And, and yeah, okay.
That, that's a whole, a whole thing.
Are we using performance enhancing
Joe: AI?
You know?
Yeah.
Yeah, exactly.
So, yeah.
But I would say for your example,
like I use, I use chat GPT for.
Ideation all the time because I'm just
me and that's what it's great for, right?
Because it has all of this
information at your fingertips.
So I think that that example is great.
You're essentially in training
mode at that point, right?
If we're going to beat
this analogy to death.
Alastair: Then the next step there
is, so we're using it for, we're
using it for ideas and then are we.
Are we using it to, uh, what's the word?
Um, so, so if, if we're using it, like
we're saying, Hey, oh, I forgot that.
Please, please add that to the
list and write a section on it.
And at that point, you're saying
it's moving out of ideation
into writing it for you.
Joe: Right.
Right.
Because now it's thought of it
and is doing the work for you.
Right.
If it's just like, Oh, I forgot
that like, Oh, live shows is a
way to make money with podcasts.
Yeah.
But, and then I say, write a section
on live, that's not like, that's not me
Alastair: anymore.
If, if you know that, if you know,
live shows inside out and you get it
to write that and they say, Oh, you
know, and this is, this is actually
my point is ultimately it should be
about the human doing the QA, the
quality assurance on the products.
On the output and if we have the
human doing the quality assurance,
the key way on everything that it
outputs and that human being signs
off and says, yes, I, you know, I put
my stamp of approval by by publishing
this or doing whatever I do then.
Like, surely that's, you've written
it because you, you, you have brought
most of the ideas to the table, you've
edited it using whatever tool you
used, be it, you know, a text editor
or a magnet to use the ones and zeros
manually, if you really want to do
that, you know, but you've, you've
edited it in some way and then you've,
uh, you've, you've QA'd it by going
over it and then you've published it.
So, but, but I think, where's the,
Joe: where's the line there?
I, so I think at that point, right?
Cause are, so let me ask you, are,
are your books all self published?
Alastair: Uh, I have, uh, I
have one book that is hybrid.
One book is commercially published
and the rest are self published.
Joe: Okay, cool.
So, so, you know, the commercial
publishing process, which is author
writes the book, send it to copy editor.
Or let's say prime editor one.
If it's a technical book, then
it goes to the technical editor.
Then it goes back to prime
editor and back to you.
Then it goes to a second copy editor.
The process you just described, in my
mind, removes the author completely.
ChatGPT wrote it.
We're sending it to an editor.
We're sending it to a tech editor.
And then we
Alastair: have a book.
Yeah, I, I, I didn't
remove it in, in my head.
I didn't remove it because the
person writing it was the person
who gave those initial ideas.
Just like in your example where.
You, you wrote it and
your editor fixed it.
Joe: But, but like if, I mean,
if I just say chat GPT, write a
chapter and then I give it a quick
once over and send it to my editor.
Alastair: So what if it's
not a quick once over?
What if, what if you then
say, uh, okay, section.
8 should be removed.
We don't need those.
Um, you missed out on this topic entirely.
Please add that into the list.
And then I don't agree with
what you said on, on section 1.
4.
So please update it to say this.
And by the way, that's a very realistic
sounding example of what I've done with.
That kind of I put,
Joe: I think at that point you might
as well just write it yourself.
Alastair: Well, you see, my argument
is no, you shouldn't because then like
that, that's, that's like going to
the farmer and saying, Hey, get out
of that tractor and here's a spade
and do your entire field manually.
It's so much quicker not to do that.
Joe: I think it's, I think, I
think that's a different knowledge.
This is the hard part of what we're doing
with our real world analogies, right?
Knowledge work.
Is very different from manual
labor with knowledge work.
It's, it is our knowledge
that is the work.
And so like by saying,
chat, GBT, write something.
Oh, I don't like these things.
Change that.
I still don't think the human did writing.
They just kind of said,
do, do that, right?
It's the difference between the
Alastair: visionary and the worker.
It's it's my argument is maybe the human
didn't write it, but they might as well
have, because the output, the quality
of the output is going to be the same.
And, and I
Joe: don't think the, I don't
feel like the quality of the
output is the same, right?
Because you can't tell chat GPT,
Hey, um, to drive home my point here.
I want to tell a story about how,
um, I refuse to record a podcast with
somebody who wasn't wearing headphones.
Tell that story, like I would tell it.
Chat GPT could never do that.
Right?
Alastair: So I think, like, Unless
you give it the details of the story.
Unless
Joe: you write the story for it.
Alastair: Well, you tell
it the story, right?
Sure.
Joe: You write it and then Joe,
Alastair: this is gonna be super
long if we don't agree to disagree.
Joe: We're gonna, yeah, we, we
Respectfully agree to disagree
Alastair: here.
Yeah.
I'm going to bring up a comment.
Uh, we have a comment on a LinkedIn.
Brian, I use chat GPT to research
the fundamentals of my profession,
which is management consulting.
In my experience, most managers have
never mastered the fundamentals.
Much like many people, I include
myself who never measured
the fundamentals of golf.
Um, yeah, so, uh, Researching
fundamentals and researching and learning.
So learning and ideation are, are, are
something that we, we certainly can agree
on that we should be using it for that.
Absolutely.
Absolutely.
I think, I think where,
where we differ is the level.
And I like the term assistive AI,
uh, because that's the way I think
that we should be using, we should
be using it to the heavy lifting,
like in, in the terrible analogies.
And just like all analogies,
no, no analogy is perfect.
But I think that, uh, you know, some
analogies are better than others.
And it's hard to find analogies for
this just because of what it is.
But, you know, um, yeah.
The way that I see it, we should be
using it to assist us and not replace us.
And I think that it's going to be
a very long time before the ideas
that chat GPT can come up with.
Or whatever AI any generative AI comes up
with on its own from it's it's learning
data set, whatever it's trained on are
going to match what an expert will come
up with what it will do is help us to
reformat and read and edit and tweak and,
um, take a blog post and, Hey, I wrote
a, I wrote a blog post on this and I want
to fit it into the middle of this book.
Can you tell me how to, uh, You
know, like, what do I need to add?
What do I need to remove?
So that this blog post can, can
then become a chapter of this book.
Here's the rest of the
book, something like that.
I feel again, I feel, well, you wrote the
blog, the blog post, you wrote the book.
All you're doing is you're
using it as a very smart editor.
I think in that scenario that,
that, that it's absolutely
fine to say you wrote that.
That's, that, that's my perspective on
Joe: that.
Yeah, I think, I think we agree
on use it for a very smart
editor where we disagree is.
It's how hands on either the
editor is or how hands on we are.
Um, yeah, I, I really think
that's where we differ.
And for me, it's, it's really about if
you're, if you're looking for me, it's
really about the, the, the personal
connections and experiences, right?
I, we can, I'm very confident that
we'll never be able to tell AI unless
there's a little, unless the humane
pin is becomes a little brain chip.
Right?
Um, Which I just scared myself, um,
we'll never be able to say, Hey,
here's all of my personal experience.
Now tell a story the way I, I
think that it's still, that's still
something that is uniquely human.
Alastair: Yeah.
And, and like, we shouldn't want to
Joe: do that.
Right.
Yeah, absolutely.
Right.
That's.
Yeah, that goes back to our,
when do we lose our humanity?
Right.
So
Alastair: who got, this has been a,
has been a fun one and it's a bit deep.
Um, Joe, we're recording this.
We're live streaming.
We're recording this for my podcast,
the recognized authority and for
your podcast, how I built this.
How I built it.
Um, can you, can you tell listeners
and viewers where to find you?
Joe: Yeah.
So you can find all of my
writing over at casabona.
org.
And of course my podcast
is called how I built it.
Um, and you can find that
wherever podcasts are
Alastair: found.
Awesome.
And you can find me at
the recognized authority.
com.
Thanks for tuning in.
Joe: Thanks so much, everybody.
Thanks Alistair.
Alastair: Thanks Joe.
Joe: All right.
I really hope you enjoyed
that ad free bonus episode.
Of how I built it.
Slash the recognized authority.
Uh, you can find
allister@therecognizedauthority.com.
I'll put that in the description
for this episode, and I'd
love to hear what you think.
Let me know your thoughts on AI in
writing what you thought about my
arguments and Alistair's arguments.
You can find me on most social
media at J Casabona, or you
can email me joe@casabona.org,
but that's it for this episode.
Thanks so much for listening.
And until next time.
I get out there and build something.