5 hours ago
8
I never thought I’d have to write these words but here I am: my name is Peter and I am human.
What seems like a self-evident proclamation needs to be made now because the misuse of AI is transforming considered op-eds such as this into “slop-inion” that is infecting the editorial pages of reputable media outlets.
In recent weeks Crikey has had to remove a series on leadership, while the features editor at Capital Brief took to LinkedIn bemoaning the fact that 80-90% of all submissions appear to be AI-generated.
Of course, plagiarism has always been a journalistic sin, and if one holds out the work of ChatGPT as one’s own, then that is clearly crossing a fundamental ethical line. But it’s not enough to run an AI check over the final copy for telltale bot-speak: the TEDx-style false negatives; the rhetorical questions, the inspirational pivot, the em dash. There are lots of grey areas in between. What if AI does the core research? Suggests the angle? Spots a logical inconsistency? When does the output stop being human?
As a loud and proud AI sceptic, I have been resistant to using the technology in my work. But recognising that I need to know thy enemy, I’ve spent the last month trying Anthropic’s Claude to understand how it might “support” my writing process.
To guide me on this journey, I’ve been taking advice from former Australian chief scientist Alan Finkel who has launched a global certification process for creators to “verify” their work is human-authored. Indeed, I’m delighted to become their first “Proudly Human” columnist, joining a growing list of authors, musicians and publishers who have gone through the same formal accreditation process.
Proudly Human applies “de minimis” to verify human authorship, a legal principle that has been used to establish inputs that would not undermine a creator’s right to copyright their work. The principle sets the bounds for what can be used to assist a creator’s work without it ceasing to be theirs. For example, it may be reasonable to use AI tools to check spelling and grammar and generate ideas. But it draws the line at drafting text or generating content that “meaningfully contributes to the final work”.
Additionally, Proudly Human recommends a more detailed series of prompts that ensure a writer is not automating the basics of story design and creative content. What we are talking about here is “provenance”, a guarantee it is me who is communicating with you through these words. A similar concept has been used to authenticate the creation of First Nations’ art in the context of exploitative appropriation.
Within these constraints I was interested in how an AI assistant could augment my current writing process. Two weeks out from filing my monthly Guardian Essential column, I will survey the political landscape and frame up some questions for the report that will tap into the national mood. I will watch, read and listen to any bits of culture that grab my attention in my search for a metaphor to torture.
Over the fortnight I will start building an argument and take a sneaky peek at the early findings to ensure the poll will fit the thesis (sometimes when it doesn’t I’m forced to throw all the cards in the air and it makes for the best columns). Then I bounce the drafts past trusted eyes (including my wife who knows my bullshit better than anyone), construct a narrative around the findings, give it a final polish and send it through to Guardian Australia editors who run their critical eyes over it.
Working within a de mininis framework I could fast-track a lot of this work, get the AI assistant to sense-check my thesis, look for linkages in the field data and even suggest some pretty lame cultural allusions. Once I had done the draft I could load it into Claude and ask whether there were holes in my logic, which it would serve up with certitude in real time. Thus, you could argue that with the support of an overly eager research assistant and an instinctively compliant sense checker, I was a little more productive.
But I also found the removal of intellectual friction took something away from the end product. For me the drafting process is critical; the problem-solving, the self-doubt, the sense that by the fourth version the piece is coming together. Because writing is more than just filling a screen with words – it’s all these choices and conscious decisions, it’s killing your darlings, it’s coming back to the well and looking for the right words to bring an idea to life.
In a world of slop, it’s critical to differentiate the work of human creators from the output of machines. A human “certification” serves both as a statement of commitment and also a guardrail from temptation should I get stuck by the dreaded writers’ block and seek an easy way out of my troubles. Because I actually don’t want to make my work easier. The hard bits are the point; the bits I get wrong are just as interesting as the bits I nail; taking the easy way is akin to the setting up the old pianola: it sounds pretty good but I’m just pumping my feet.
For those of us who are concerned about the impacts of AI, we need to do more than shrug our shoulders or shake our heads; we make choices every day. An authenticator such as Proudly Human is not just a commitment from creators, it is also a proof of life that consumers should demand if they really care about the sort of society that comes out the other end of this so-called revolution.
Cultural content is the ground zero of the AI insurgency, the theft of intellectual work leading to the replacement of those same creative workers, and if we can’t draw a line here, I fear there won’t be any lines anywhere. Provenance matters.
-
Peter Lewis is an accredited Proudly Human writer. He is the executive director of Essential, a progressive strategic communications and research company, and the host of Burning Platforms podcast. No AI was used in the creation of this column
