There’s a fair chance some (or all) of this will be considered quaint and out-of-date in a pretty short time. Despite this, I still really felt the need to write down and formally document my thoughts on this topic as it is permeating more and more of my life. I don’t expect that will ease up any time soon either. I also feel that grappling with how we interact with, use, and feel about AI is potentially going to be the philosophical and ethical topic of our time.
Writing, talking, thinking
I have heard convincing metaphors on both sides of this issue. On the one hand, there are those who say things along the lines of:
AI is just a tool, just like we don’t think twice about using a lighter or a calculator, we shouldn’t feel guilty for using it as a writing tool.
On the other side are people who argue:
Writing is thinking. The process of writing is iterative as we refine and reflect on what we want to say and how we want to say it.
The piece of writing you’re currently reading is evidence of that exact process for me.
Now I’m not going to say that either of these perspectives is right or wrong. I think there is a lot of virtue signalling going on in the discourse surrounding this topic and I want to stay away from that if I can.
What I want to see if I can do is bring these two lines of reasoning together into an alternate perspective.
There is another aspect of this new technology which I think has made this merger possible: dramatically improved speech-to-text technology. In plain terms, I can now speak (let’s be honest, ramble) to a prompt box and convey my thoughts in a much shorter time. Maybe it is just me, but sometimes, when I am writing long-form prose, I have a neat idea while I’m part way through writing the last thought. By the time I finish it, the new idea has wandered off. Being able to literally speak my mind has really helped alleviate this problem.
I have certainly found that since I started “talking” to AI, I have been getting much better responses. This ultimately comes down to the AI now having more context. The throwaway statement buried in my meandering ramble turns out to be more important than I thought because it makes the implicit explicit, and this is one of the keys to getting the most out of AI in my experience.
AI as a collaborator
The next thing that “talking” to AI has unlocked for me is the feeling of a collaborator. Writing-as-thinking seems like a “cleaner” form of thinking to my mind. Following that analogy, conversation-as-thinking is a “messier” form of thinking. But this is where AI creates a new paradigm. Messy conversations with someone are great, but often when I reflect back on them, I forget some of what we spoke about, or I can’t recapture the fascinating insight we crafted. When having such “conversations” with AI, there is a full record of everything being said. I can speak with the freedom of knowing I don’t have to try and commit everything to memory, and can instead focus on the next train of thought. So AI is like a collaborator who you can talk to at any time, anywhere (with an internet connection), for as long as you like. It doesn’t judge you, it doesn’t get bored, and it doesn’t get tired (unless you hit your usage limit).
One way I, and many others, have been engaging with this new way of collaborating with AI is through skills. For myself, a powerful little skill called /grill-me has been a highly effective way of exploring this conversation-as-thinking paradigm. What this skill does is get the AI agent to ask you questions relentlessly about some idea until you both reach a shared understanding. I almost always get to the end of these sessions in a much clearer place, and quite often in a very different one from what I anticipated. Something else that happens regularly is that we get a few questions in and realise the idea falls apart because I had some flawed assumptions or a blind spot in my thinking.
I have been writing a lot of fellowship and grant applications recently, and I have found this skill immensely helpful in crafting my ideas and responses. And this brings me to one of the key points I want to emphasise:
AI doesn’t necessarily make me write faster, but it does help me create more nuanced responses than I would produce alone.
This comes back to the point that when you treat AI as an always-available collaborator, you can explore ideas without always needing to have a colleague available (and willing to listen to your crackpot early ideas). Don’t get me wrong, I still get a lot out of bouncing ideas back and forth with my colleagues, but I have found that AI has become a very useful addition to my toolbox.
Substance, not signs
I was deliberating on whether I should finish this piece by giving some kind of advice on when, and when not, to use AI for writing. But I kept coming back to this nagging thought that the process matters less than I first assumed. There is a broad spectrum of ways people can use AI to help with writing, ranging from using a single prompt to write a blog post with no to-and-fro, to the conversation-as-thinking pattern I have just described. And then there are various settings in which we write: social media, blogs, grant applications, scientific papers, strategic memos, birthday cards—the list goes on. I don’t think there is a universal rule for how people should use AI to help them write.
The more important question is: are you proud to put your name to the output? Ultimately, you are solely responsible for the output, so you had better own it.
There is a lot being asked of writers when it comes to AI use, and a fair bit of judgement. But I think it is important to remember that the reader has a responsibility too. It is easy to fall into the trap of identifying telltale signs of AI-generated writing. Tunnel vision sets in until all you do is look for the next sign, forgetting that you should be reading the substance. And look, if it is utter slop, there may not be much substance. In that case, I think instead of calling out the AI, we should be calling out the writer, who has taken full responsibility for the output by sharing it.
It is not a reader’s place to tell a writer how they should undertake the process of writing; everyone has their own way. But it is okay as a reader to demand content with substance.
So I guess this post has turned primarily into a call to readers. Focus on the substance, not the signs. You might be missing out on some really interesting ideas if you don’t. And if you are a writer, I hope this post has given you some food for thought on how you might use AI to help you write (and think).
I want to end with a quote from a recent episode of The AI Daily Brief.
The purity test for AI writing will die.
The quality test won’t.