763: Workflows with William Gallagher

Episodes
21

Oh, I thought you were using AI to help you thinking through stuff before you started writing.

Maybe you should give it a go, sometimes? You could still write your first draft by hand. The too-and-fro chat with AI trying to understand a gnarly problem is sometimes fascinating and enlightening. It’s like cocreating with a clever friend, who sounds confident, but doesn’t understand your domain.

Let me give you an example:

I have a model called the AgileTocMethod, and at the heart of it is what I call the thrive model. I’ve spent hundreds of hours and thousands of dollars working it through over the last 3 years, pre AI. In the last week I decided to work on version 2, because there was something wrong with it.

And I needed someone to talk things through, so I’m using ChatGPT. So far I’ve spent about 5 hours using it to think through my tweaks.

There are less than 50 words on the page, and they’re in a venn:

IMG_0016
IMG_00161679×1250 168 KB

On Thursday, I spent an hour talking through the diagram with ChatGPT, describing it as a venn, and labelling each bit. This was useful for me.

I then told it I was going to tell it about each bit, which I did - I went into voice mode and dictated it. Very garbled! And ChatGPT took my garbled words and kinda cleaned them up, but it often misinterpreted my words, so I explained what was wrong and together we fixed them. (Just like in a conversation with a human, like ours).

I asked it to summarise what I’d told it, using the labels on the venn as headings, which it did.

At this stage I had, for the first time, a nice summary of my model.

I am now working through each section, rereading what ChatGPT and I wrote, correcting the words and meaning, adding in a few meraphor. As part of my upgrade, I decided to change some of the labels and it helped me come up with new names. For instance, I changed the right hand circle to Leveraged Leaders, which changed the meaning of the intersecting in the bottom. We chatted about what you’d get when you had great leaders and you could also deliver projects on time, and it suggested trust, which what I was thinking, and I asked it for a 2 word name, and it came up with enduring trust. And then I changed leveraged leaders to trusted leaders.

Soon I will have a 3-4 page “shitty first draft” white-paper, which I can edit, and every single idea in it is mine. And, every single word.

Sorry, that was long Jim. I’ve read your blog and I suspect you’d love that kind of thinking. You don’t have to use any of the words from the chat, but the conversation can be incredibly useful,a] and enlightening.

For me the mental sparring and cocreating with ChatGPT is like magic.

could hire someone to help me do that, but I get all that lovely, creative too-and-fro for $5US a month!

1 Like
22

You’re an adventurous man Barrett, and an experimenter. I know you don’t feel comfortable using ai to write drafts, and nor did I until about a month ago, but I wonder if you could do a few experiments using ChatGPT to write a first draft, and tinker until you feel comfortable?

The reason for my discomfort was that I kept reading stuff about using ChatGPT to produce massive amounts of cheap content.

Like horrible large scale chicken farm/factory’s churning out millions of words/chickens to feed the masses and make lots of money.

They can do that, if that’s their job, I guess, but I cannot.

Just thinking about it makes me :face_vomiting:.

So, I don’t use ChatGPT like that. It took me a while, but I now use it to, first, take my original thoughts and tease them out. I produce a first draft by too-ing and fro-ing with chatGPT. And then I edit it by hand.

The too-ing and fro-ing is the bit I have had to get comfortable with. ChatGPT spits out very average words, and it’s often in passive voice, and its writing is fast but bland. I’m finding I can ask it to write differently and betterly, and it does a lot better job than I can on my own.

(Btw. Just in case it’s not clear, I’m still learning this process and I write most things, like this note, by hand. You can probably tell by the typos…)

23

What you’re describing is still editing, not writing. They are separate skills (and have been defined separately for several centuries). If you swapped the LLM for a human and bounced ideas with them and asked them to produce a first draft for you based on your discussion, you would presumably not claim credit for that writing (and if you would that’s a whole other problem).

Personally, if I learnt that someone whose thinking I valued was doing that, I wouldn’t be happy. To me it shows an error in judgement. Learning how to articulate your ideas and how to write well are both skills, and they’re valuable skills. By using an LLM to do your writing (and then taking credit), you’re bypassing the practice needed to develop those skills. And as others in this thread have noted, the writing process is part of the thinking process, so you’re also cutting out a valuable step where you might have nurtured other ideas.

There’s also a separate but equally important point that I wasn’t going to make until you remarked on the cost: if you don’t value the skill of a good writer, why should they value your skill as a “thinker/editor”?

Creators who use AI for one skill but expect people to pay for the equally replicable skill they’re offering are exceptionally hypocritical (that’s not directed at you specifically, just in general). I’m seeing this play out with a lot of newsletter writers at the moment - they expect people to pay for their writing, but they’re using imagery generated by Midjourney, etc. If you believe your skill should be compensated fairly but are happy to exploit someone else’s skill for your own gain, it shows to me you don’t value humans or skills generally, and I’m not interested in engaging with that.

3 Likes
24

With genuine respect :man_bowing: I disagree with the propriety of having AI generate the draft. If AI produces the draft for me, it is not my writing. I’ll grant that the more you change the original AI draft, the more the text becomes yours. Nevertheless, if the “guts” of the text were AI-generated, it is not authentically that of the author. It is in many respect, artificial. I may or may not be adventurous, but I have ethical qualms about posting an article under my name that was essentially AI-generated, no matter how many edits I make. It is not my work; it is an edited amalgamation of text scrapped from the work of others.

As to “Like horrible large scale chicken farm/factory’s churning out millions of words/chickens to feed the masses and make lots of money,” it is not quite that easy. Let me start by making it clear that I believe we have an ethical obligation to be as humane in our treatment of animals as feasible. We should not make the lives of animals miserable. That said, when one has plenty of food on the table, it is easy to condemn mass food production. If one is constantly hungry and can’t feed one’s family, mass-produced food that is cheaper than it would otherwise be is essential. Food production methods, like environmental issues, are more complex than virtue signaling (I’m NOT accusing you of that) suggests. Any policy must consider its effects on everyone, including the impact on job creation and income for the poor. My point is not to discuss food production or the environment; it is only to point out that most public policy issues are more complicated, with unintended consequences that many do not take the time to consider.

Just my two-cents. :slightly_smiling_face:

4 Likes
25

I’m quite interested in this discussion about the use of AI in writing–particularly the when’s and where’s. This morning’s edition of Total Annarchy (from Ann Handley of Marketing Profs) was on point, IMO, and I’m happy to see that only 5% of survey respondents use AI for a first draft.

3 Likes
26

Indeed. I particularly appreciated this:

We have it backwards.

The first draft is the thinking draft. That’s where you need to be fully present. On board. Just you and your glorious ideas.

Otherwise, you undermine yourself. You shortchange your growth and creativity. You put the ROBOTs first. You put yourself second.

The first draft is the thinking draft. (Said twice for emphasis.) Protect the part of you that truly makes your content, your writing, your voice your own.

Why? Because otherwise and over time, your skills will atrophy. You will doubt yourself more. You will build a dependency.

We will begin to rely on the Really On Board with Outstanding Tech (those ROBOTs!) in exactly the way the tech companies want us to. In exactly the way the lazy part of us is willing to.

Again, I’m talking first drafts here.

Gen AI is useful in many situations. I use it. It’s helpful. But not in the actual getting ourselves onto the page part.

4 Likes
27

And much, but not all, the revising is ALSO thinking. You realize that this point is related to that point and so those two sections should appear consecutively. You realize this other thing has nothing to do with the main point — so cut it out.

As you near the end of the process, you reach a point where it’s helpful to feed the draft to AI and let it point out spelling, grammar, usage and punctuation errors, and places where you can trim words for tighter prose.

1 Like
28

Oh, that’s interesting …

Maybe I don’t know what writing and editing are? You’ve got me thinking …

What do you think writing is? Is it the typing of words?
And what makes it different to editing?

1 Like
29

Same re the mutual respect, Barrat.

I don’t currently subscribe (ha ha!) to your view, so it’s good to chat here.

I have a dilemma!

I don’t know if you saw the example of my ATM model earlier - the one with the Venn diagram.

I have spent, I guess, two hundred hours putting the diagram together (all 50 words - 4 words per hour!) and I’ve given talks on it and written articles about it. And I’ve used it for the last 3 years to sell and deliver consulting and mentoring work.

It sums up the work I’ve done since I completed my mba dissertation over 20 years ago.

Ive got 3 books out there, and my work has made the working lives of hundreds of thousands of people happier and richer.

But here’s where I’m stuck.

I don’t have a nice 3-6 page description of the model.

So, in the last few days, I spent about 6 hours describing the model, and talking it through with ChatGPT, and also, sometimes, Claude.

I’ve dictated most of it, typed a little bit of it, but all my words are now in chatGPT, and under each heading we’ve worked out a couple of paragraphs. Each paragraph has been hard work. I talk a bit, ChatGPT talks back, we too and fro, the words start to form, I ask it to rewrite in active voice. I tweak stuff. And so on. Some times I copy a paragraph, edit it, then give it back to ChatGPT, and it makes a few suggestions. And so on. Very iterative.

I’ve got another 2 hours to go and I’m delighted with my work.

Really delighted!

This afternoon (it’s Monday here in nz already) I was going to ask ChatGPT to write out everything we’ve worked on, using the labels from the ATM venn, and that’s be my first draft. I was then going to copy it into scrivener and rework.

It would take about a minute for chatGPT to spit out the draft.

But now - seriously - I’m wondering if I should type it out by hand?

And, also, I’m wondering if those words are mine?

—

I know what I’m going to do, but I worry that maybe there’s something morally wrong with asking it to produce my “shitty first draft”?

30

If I were to take someone else’s writing, rearrange it and change some of the words, when would it stop being plagiarism?

31

Maybe that’s the fundamental question here, Karl.

Perhaps there is a continuum? It’s definitely not black and white.

I’ve got a good example of bad:

Someone downloaded one of my books last year, changed the cover, gave it a new name, and started selling it on Amazon.

That’s clearly theft.

A different example:

Other people copy bits of the same book and share them on social media.

That’s flattering.

A senior advisor in the UK government has just taken the same book, broken it down and created training from it which (may hopefully) be used to reduce hospital waiting times.

That’s so cool!

Someone else in the US Airforce did something similar and changed the wording in one of my frameworks. And … honestly they made it better.

That’s awesome too. Especially since they credited me. We’ve become good friends.

And all of the ideas from that book were my simplification and modernisation of a book that came out in the 1970s.

1 Like
32

Oh, I see, plagiarism is a continuum, eh? Giving credit, or fair use, or getting at least tacit approval (even after the fact) seems to be enough to make it alright with you.

33

Oh, I see, plagiarism is a continuum, eh? Giving credit, or fair use, or getting at least tacit approval (even after the fact) seems to be enough to make it alright with you.

No - what makes you think that?

Plagiarism is at one end - the bad end - of a continuum.

Does that make sense?

1 Like
34

I specifically asked, " … when would it stop being plagiarism?" and you did not answer that question. Your many examples did not show me where you draw the line at plagiarism.

You have said that someone else writes your first drafts. To paraphrase my earlier question, how much rearranging and changing of words must there be before it stops being plagiarism?

35

Ah … words can be so confusing.

I did the consultant thing, and “bounced off” your question, because I suspect all of these discussions boil down to a continuum.

Why do you write “You have said that someone else writes your first drafts”?

I have hired a copyeditor in the past to write marketing stuff for me. And I use editors on my books to improve my words. But I don’t use anyone else to write my first drafts.

1 Like
36

Thanks for sharing the excellent and thoughtful article, Cindy.

I agree with the author that writing a first draft is thinking and also expresses the writer’s unique way of communicating or voice. However, you lose all of that by letting AI produce the first draft.

Many who use AI extensively seem to assume that those who don’t haven’t experimented with it much, and if they did would likely use it more. I’ve experimented with ChatGPT a lot. It’s made me more aware of how easy it is to allow AI to supplant the human creative process. It’s helped me to define what I don’t want to use AI for.

Just to see what would happen, I tried feeding AI my basic ideas and asking it to generate a first draft. The results were impressive, but they weren’t me. I tried doing “heavy editing,” but I found that the existence of the AI-generated text too strongly influenced my sentence structure and the way I would say things. The experiment resulted in me being the editor for an AI-generated blog post. I was not a writer; I was not a creator. I was not “collaborating” with AI to create.

That experiment demonstrated to me the dangers of using AI to generate a first draft and how it destroyed my unique human creation. Thus, I have a rule that I do not have AI write my first draft, no matter how “helpful” or “time-saving” it may appear to be.

37

It sounds like you’ve found what works for you Jim.

I hope my notes the other day didn’t annoy you.

Our experiences differ, that’s all.

1 Like
38

With respect, I disagree.

Our differences are not just based on “differing experiences.” We have significant value, philosophical, and definitional differences about what it means to be a unique human creator and writer.

39

Ouch.

I don’t think we do, and I’m disappointed that you do.

2 Likes
40

I’m going to step out of this thread since (what I thought would be helpful and delight everyone) seems to be annoying people. And it’s making me feel kinda stressed, which I don’t need. Sorry about that.