I noticed that last week Substack introduced a rule requiring authors on the site to disclose whether they use AI in the production of their work.
I have to say, I find that requirement very strange.
They do not require that the use of other technology or supporting methods be disclosed.
There is no requirement to say if paper has been used, or a pen or pencil, or of what type.
Nor is there any suggestion that the use of a dictionary should be disclosed, whether it be a bound volume or electronic.
Come to that, it is not necessary to say if a word processor has been involved in creating the words published, or whether any of the edits that it might, these days, suggest have been adopted.
Nor am I required to disclose if I use mind map software to organise my initial thoughts, or a spreadsheet to produce any chart that I include, or other software to produce diagrams that explain my ideas, although I do, almost invariably, use these tools in my work as I require them.
Come to that, no one is asking that I disclose my use of dictation software on my phone, which is what I am using to write the first draft of this piece.
Every single one of these things does, in some way, aid the creation of my published output. I do not use them all the time, but everything that I do will involve at least some of them, and quite often many of them in combination.
As a writer I have published, at a conservative estimate, at least 10 million words, and probably a great deal more, on my blog over the last 20 years, I've learned that there are many and various technologies that can assist me to get my words in front of a reader, not least including internet-based publishing software of WordPress, Substack and others, of which software I have little true understanding. But I am now required to disclose my use of AI.
My question in that case is why, when so many technologies are involved? Why, too, should I be required to disclose my use of AI when I almost entirely restrict it to searching only those materials that I have already written and to using those alone to support the arguments that I am now presenting afresh? The exceptions are when I use it as a better search engine than those previously available, but for a quarter of a century or more now I have not been required to disclose search engine use. What, then, has changed?
In particular, why do I have to make disclosure when my style of use means that I very largely use AI to save myself the time of re-creating a wheel I've already constructed? Using it that way, I can check that it has not made it square rather than round, because the source material it uses should always have been of my own creation. Why I need to disclose this appears particularly odd to me, but it would appear that this is what Substack demands. If, as I presume acceptable, I use it to edit and repurpose my own copyright material, I cannot see why I need to make any such disclosure, any more than I think I need to disclose the fact that I use AI as a spellchecker, because almost everyone now does.
There is, though, one exception to this. Having written the text of the AI infographics I have been publishing of late, I use that material to then create those presentations. I can no more create them without AI assistance than I create the many charts I have created in Excel over the years.
That said, the written content of these is always developed in my usual fashion. The process starts with either a dictated note or a mind map. Because the eventual planned structure of the output is now well known to me, it moves from the initial thought collection process into a series of structured bullet points quite quickly in many cases. There are, of course, exceptions. Because I impose a constraint upon myself, which is that these pieces must, in my opinion, be comprehensible to a curious 12-year-old, I sometimes have to work quite hard to achieve that goal. The infographic on central banking did, for example, probably go through 15 drafts before I got to the point where I wanted to try making a graphic. That, however, is exceptional. Five versions is more typical. Most of the time I write about topics where I have a fairly clear understanding of precisely where my work will be going. And always, and throughout, I remain in command of the process, even when. I ask the opinion of others, which I very often do. The mistakes are all mine.
It is only when I'm happy with the draft text that I make the graphic. That is the last point in this process, and the task that I could not do. That said, I cannot create digital photographs, and I use a camera and software to do that. Nor can I produce flow diagrams, but I have software that does. Similarly, I have electronic mind-mapping software, which I can use to produce visuals. I see no difference between those processes, which interpret my data, and the infographic that I produce using AI, but apparently Substack does.
So, I make full disclosure. I use a lot of technology when writing. Amongst the technology I use, AI can on occasion be used to find, organise and help edit my work into new pieces that I am writing on themes I have previously addressed, but which I wish to revisit. And I do use AI to present some of my work, just as I use the software in WordPress, Word, PDFs, Substack and the other programs that I've already mentioned to sometimes do so as well.
Does any of that suggest that my work is not original? I think not.
I am not naive, though. I am, of course, aware that some people can use AI to produce output into which they put little effort, and even less understanding. I do, however, think we need to be a lot more discriminating when we come to AI disclosure. If it is just another tool that the writer uses to organise their thoughts, I am really not sure what the issue is that we are meant to be worried about.
However, if people want to know what I do with it, I have now explained, and I no more apologise for the use that I make of it than I apologise for the use of my time, my thinking processes, and all the other technologies that have let me put words on screens for more than two decades.
Thanks for reading this post.
You can share this post on social media of your choice by clicking these icons:
There are links to this blog's glossary in the above post that explain technical terms used in it. Follow them for more explanations.
You can subscribe to this blog's daily email here.
And if you would like to support this blog you can, here:
Buy me a coffee!
I can understand your frustration at being asked to disclose one particular tool you use to do your job. In your case I fully agree it is unnecessary.
However …
Yesterday I noted a YouTube video apparently by Roger Penrose. He’s a Nobel Prize winning physicist whom I find interesting. As I started to watch him present, the language seemed stilted but the image and lip sync were good. It’s was only when he started referring to himself in the third person I was certain this was AI generated slop.
It turned out to be essentially impossible to report this to YouTube. This was a whole channel with dozens of similar videos all pretending to be by Roger Penrose. And there is no way to report this.
The European union is introducing legislation to require that AI generation is disclosed. So are, or have, South Korea and some US states. In my opinion this is necessary to prevent social media being overrun by AI slop.
Unfortunately I think you are being caught up in this. I hope that there will be some way to disclose whether content is “AI assisted”, which is what you do and is perfectly legitimate, as opposed to AI generated, i.e. “slop” which is not acceptable.
A good distinctionm.
I use AI as a tool, like I use many other tools, including this computer right now.
[…] In some way or other, I used AI in the creation of this post. I disclose how I use that technology here. […]
“The European union is introducing legislation to require that AI generation is disclosed”, how will they or indeed substack know?
It is perhaps worth making a distinction between the use of AI from time to time and the use of a trained AI. I use a trained AI, trained on a range of docs (my docs) plus the various discussions we have had (most of which would give many people pause for thought) . So not your “common or garden” AI.
One particular area where AI is good (at least in my field of interest – energy) is analysing the nonsense the EU regularly produces: “Actions Plans”, “Impact Assessments”, “Directives”, “Regulations” etc etc ad nauseam. I have reached the point where – apart from speed reading some of the stuff – I just pass it to the AI for analysis. I point blank refuse to read a 400 page Impact Assessment (typical length). The downside for the EU is that within 10 mins of publication of one of their “excretions” I can have a 3 or 4 page tightly worded critique (suitably “de-AIed”). Doubtless this is very irritating for them, I hope.
Thanks
BT many years ago produced software to make a synopsis of scientific papers. It claimed something like 90% accuracy from a synopsis of just 20% of the original. “Why use one word when ten will do?” Joke!
Pretty much what Tim Kent has said really. There’s been particular public backlash against the use of AI to generate assets for videos, artwork, even placeholders in video games (Divinity game) where the majority opinion is that this has replaced an employee or a hire. You say that you cannot have made the infographic but another human could have. To many people they will be critical of your use of AI in this regard because it’s an artist of some description who might not have work and it’s very likely the AI model you used effectively stole the content from someone ho made infographics for learning to reproduce. This is not my opinion but one many of my friends hold which is not without merit and does cause them to look elsewhere.
I do agree there should be distinction in what AI is used for ideally because it will be used more widely in the next decade.
While we’re talking about AI and generation, It’s worth pointing out your video style makes it very easy to train an AI model for video generation. While I wouldn’t advocate for someone to do so, they could very easily generate fakes of your videos largely indistinguishable criticising MMT and professing the triumph of neo-liberal economics.
I could ask a scribe to write with a quill pen for me too.
I could also type on a manual typewriter.
I could ask somneone to run on a treadmill to generate my electricity.
Shall we get real here?
And should I stop producing video because they might be copied (and any format can be now).
What do you want me to do? Give up? Be like Socrates and not write a word down? Or live in the real world?
What was your aim in writing that comment?
Richard,
I think Alex is not argueing with you saying you use ML/AI as a tool. Rather he is pointing out that your tool has been trained to make (in your case) infographics on the work of real illustrators. These people have not been paid for the copying of their work nor are they being hired by you to make new work. This is unfair in my opinion and might be something you consider as a champion of a fairer society.
Just my pennysworth as a graphic designer turned carpenter (to spend less time behind a screen).
Cheers, Andrew
I know what I am being told, and it makes no sense.
First, this demand would require that the graphic designer never copy the work of others, and they all do.
Second, it means I should letterpress typeset this blog or else I take away the jobs of those who once undertook that trade.
Third, it means no one should take a photo – as that takes away the work of a portrait artist.
These arguments are just ridiculous. I live in the real world. Tech changes. Argue about power use, by all means, but not jobs changing. They always have. The question is how we deal with the bigger consequences of that, and fixing the world in aspic so nothing changes is no solution, so why promote it?
I’ve been reviewing publishing agents and publishers for a project. They almost all now have a stern warning that they can recognise AI generated prose and will reject any submissions containing it. One reason given is that AI relies on published work but neither respects nor acknowledges authors’ copyright in their work. However, most of them also acknowledge the legitimate use of AI in research. It seems to me Substack is making a spurious conflation of AI “slop” with plagiarism – two entirely different things.
I can’t plagiarise myself….
Substack is using Pangram, a neural network called a classifier model, to scan posts, notes and comments for evidence of AI use in writing them. In effect, using AI to detect AI. Pangram the company has its own Substack newsletter. The comments beneath its post describing how it is helping Substack are almost entirely negative. Few people on Substack seem to want it anywhere near their writing.
Thanks, Ralph.
The paradox is obvious.
Pen, pencil, quill and paper, a typewriter, a word processor, a printer, all are fully under your control. AI is not. The output AI generates depend on the data AI was trained on as well as how it was programmed to respond to your input. Changing the order of the input may result in different output, maybe carrying the same meaning, maybe a subtle, or even a blatant different meaning.
Do you think I do not raad every word?
I said I check evertything.
Please do not treat me like an imbecile.
I think you have made a mistake of putting yourself and your personal experience at the centre of the request, instead of thinking that the disclosure might be useful and be an important guardrail. The ‘genie is out of the bottle’, but you shouldn’t be ignorant of the fact that the labour of people was stolen in order for you to be able to use AI; it might be free and convenient for you, but someone else has paid the price; they have lost their livelihood, have poor water quality, massive power bills, the speculation secondary markets. AI makes it easy for people to distort what is real, disclosing AI isn’t about you, it’s about the people abusing it. There are a lot of things in this world that are bad for society that we can’t disentangle from our lives, but we shouldn’t lie to ourselves about the human cost, nor complain when asked to provide a little transparency. You would be correct in saying you are using the ‘technology’ if you built or trained your own LLM (but you didn’t?), so please don’t conflate using an ‘AI’ service with the technology itself (LLM).
I am sorry, but I think you’re wrong. It’s just a tool. So was a word processor 40 years ago. It transformed work in offices. The secretary disappeared. Are you saying we should go back to manual typewriters. Why not? I strongly suspect you are being very inconsistent. And you won’t convince me.
Dear Richard,
Once again you reply using arguments that do not answer the point.
Nobody
is suggesting quills, typewriters, no more photos and these are cheap shots. AI steals from creative people and using it supports that. Everyone is free to use as a tool but denying how it ‘learns’ is not an valid option.
I am afraid that you are increasingly unable to take comments without feeling there is a personal attack. For me this is now a reason to stop reading your daily writing, which makes me sad as I have learnt so much from you over the years.
Wishing you health and wisdom, Andrew
The sad things is, you have not engaged with the points I raised.
It is quiet absurd to argue this is the first time creative people have lost their jobs to tech.
Feel free to do. It would be more use to really engage.
[…] In some way or other, I used AI in the creation of this post. I disclose how I use that technology here. […]