Real Artists Ship
Reflections on ChatGPT and Deep Thinking
Fragmentation is the enemy of all that is good —
or so it feels, a few weeks into using ChatGPT (and about a lifetime of using computers). I first used it the day after it was launched, and I was blown away. It became clear to me that the future of Computing was changed forever. It helped make tangible the promise of AI in a way that I just wasn't quite able to grasp before.
On the first day, I spent about 3 hours. On the second day, it was more like 8. The day after that, I even woke up in the middle of the night to chat some more.
And then... something interesting happened. Looking back, it feels like my insights were not really going anywhere. I had these conversations with it, and it helped me engage with topics that I care about in a completely new and thought-provoking way.
But, at the end of the day, the chat would end (this was before it maintained your chat history), and the context that was built up would be lost. It felt like I was putting in careful work to build up shared understanding, and then it evaporated without any artifacts to show for it.
Something else happened, too. I would have a new idea. And instead of writing an explorative essay about it (which I'm just now realizing is a thing that works extremely well for me, and something I don't do nearly enough of), I would chat about it with ChatGPT. And it would naturally engage with my thinking in a way that produces shortcuts to some kind of insights.
But my overwhelming feeling now is that it is not nearly intelligent enough to consistently reach the kind of deep understanding that is required for truly new and interesting ideas to emerge. It doesn't have the judgement required to identify really great ideas.
And more importantly, it somehow shallows my engagement with new ideas to the point that it actually adds more Noise rather than helping me focus on the Signal.
I often take long breaths after someone asks me a good question. And I often find that the pace of conversations - personal, intellectual, and strategic ones - is too quick to actually think about things at adequate depth.
Like, most of the time, for most things, I'm not interested in just reproducing known answers. I want to get somewhere new. To realize something more interesting than what I already know. And that requires a certain degree of slowness and depth — because that helps me attune to the thing in question, to find the signal in the noise.
Now, in my conversations with ChatGPT, it increasingly felt like I was speeding things up. It responds so quickly, and engaging with it is very thought-provoking but on a somewhat shallow level.
By leading with outputs rather than questions, it's adding words to the conversation, without necessarily deepening the engagement with it, without picking up on the signals of interestingness, and without slowly picking up on the threads that might lead somewhere profound.
Now, the obvious disclaimer is that all of this might be completely changed when we get to play with GPT-4 in a couple of months — in this case, consider this a kind of witness testimony in an early moment of quickly unfolding history.
But actually, I don't think so. I think these are observations about the medium, rather than the quality of it's messages.
(Pun previously not realized, but now fully intended).
Makes me think I should write a thing titled ChatGPT: The Medium and its Messages.
And you see, the quick way of doing that would entail opening ChatGPT right now, giving it that title, and asking it to write an essay outline. Asking it to make arguments for the core claims. Etc etc.
But that would take away the joy of actually thinking about it. In fact, it would, by overwriting the blankness of the page and the vast intelligence of my intuition meeting that page (not to brag, to make a point, I swear), actively get in the way of engaging with this topic, exploring this field of questions, insights, new perspectives, ... in the deepest way possible.
So what would an adequate engagement with this topic look like?
For one, it would mean actually writing that essay and publishing it. That's the hunch that inspired the title of this piece. For me, personally, much of my best thinking happens in the context of writing something I intend to publish. In fact, this draft, so far, is an incredible example of that. It helps me make sense of this whole experience and topic much more than any of my private writing on it has so far. And maybe that's just a personality thing, but there is a kind of rigorousness, attentiveness, and explicitness that comes with writing for another mind.
This reminds me to make an important disclaimer: Understand this writing as an exploration of my personal process of thinking and thoughtfulness first, and an exploration of the principles of the medium and its messages second, and the principles and possibilities for how it can Improve Human Thought third.
This brings up another interesting point: The value of uninterrupted long-form writing. I probably would not have gotten to this paragraph I just wrote, I wouldn't have gotten to participate in the experience and creation of this very storyline that led to it — if I had tried to chat it into existence.
Words are useful and significant, and in hindsight, it's obvious that "chatting" isn't bound to lead to the deepest kind of thinking and insight. It also doesn't lead to the production of artifacts — it's an inherently ephemeral activity.
This leads to yet another insight: Meaningfulness. The way I was able to engage with my experience and thinking on the topic by just sitting down to write this essay is profound! (Mind you, after waking up at 3 am, jetlagged in an Airbnb on my fourth day in Tokyo with the intense desire to write this all up — maybe this is how my brain substitutes REM sleep.)
This open-ended style of writing, intended for publication, allowing for the surfacing and exploration of ideas, is perfect for weaving together a story. To make sense of what's been happening so far and find meaning in the questions and possibilities that follow. Like:
- What might be done about it all?
- What should be my relationship to this technology?
- How can I do my best thinking?
It also surfaces open threads (haha) in a way that chatting doesn't really do. I very much like the meta-cognition that it brings.
Real artists ship, and real thinking entails producing artifacts worth shipping. In a way, this isn't just about how to use ChatGPT. It's an inquiry into the nature of thinking, the conditions of the digital tools and environments available to us in 2023, and the behaviors, practices, patterns of attention, and thinking we default to by inhabiting them.
Here's what I mean:
The sense of fragmentation and my frustration with the setup of my second brain isn't new. Over the past year, I've significantly increased my engagement with Tools for Thoughtfulness. Right now, I'm typing on my focus-typewriter: a 12-inch 2017 MacBook with a few good tools and almost none of the distractions. (Here's the previously unpublished blog post I initially wrote when I first got it up and was very excited about using it as a tool for thought).
I'm also using: handwritten paper journals, a Remarkable, Obsidian, iA Writer, Figma, Things, Notion, Freeform, and ChatGPT, ... in other words, my thoughts are spread out everywhere, and I've been really struggling to create a system that integrates them into a coherent, enjoyable, and useful whole.
But my hunch and the core idea of this essay is this: Real Artists Ship, and writing with the intent of publication is the single most meaningful practice for good thinking.
And really, that might mean creating something like a wiki. The paradigm of networked thought has merit. If your life's knowledge work is going to expand the frontier of knowledge in a meaningful and significant way, you'll probably want to link between your insights, ideas, principles that you stumbled on, etc
"Cultivating a wiki" might also just mean writing blog posts that reference each other.
And that, I believe, will be the backbone of how I engage with the future of Computing. It's also what I'm curious about exploring to augment with LLMs — what tools, environments, and processes can support us in thinking through our most interesting questions?
In writing this, I repeatedly noticed a slight resistance to using certain words. "Augment," for example, didn't feel quite right. Neither did upgrade or facilitate — all of them are words that I previously used a lot. Maybe the lucidity of this early morning stream-of-consciousness writing helped my subconscious pick up on premises embedded in these words that are worth challenging.
In editing, I noticed how I made italic, both while writing and during my first edit pass, anything that conveys significant meaning. Some hints for future exploration: How about creating wiki pages (or an equivalent thereof) for each of these things that carry meaning? How about producing questions to help write out what's meaningful about these ideas?
What do I mean by good thinking? I mean original thinking. I mean exploring the frontier of what you know and how you understand the world. I mean finding the insight, truth, and meaning that can inform how we live our lives, what we work on, and how and why we do any of those things. Good thinking is an inevitable part of a life well lived, a life dedicated to the realization of our highest ideals, values, and purpose, a life adequately expressive of our higher potential and the beautiful possibilities of the worlds we are yet to create.
Worth noting that the phrase Real Artists Ship originates from Steve Jobs. This is a fun summary of the origin.
Steve was fond of summarizing the themes of the day into a few succinct aphorisms, which he called “Quotations from Chairman Jobs”. The sayings from the previous retreat, held in September 1982, were “It’s Not Done Until It Ships”, “Don’t Compromise!” and “The Journey Is The Reward”. This time, they were “Real Artists Ship”, “It’s Better To Be A Pirate Than Join The Navy”, and “Mac in a Book by 1986” — Credit Where Due, Andy Hertzfeld
Thank you for reading. I'm also grateful to Nick Punt, for introducing me to the world of wikis and for the quote that inspired the title of this essay, and Oliver Klingefjord, for our previous conversations on the brave new world of LLMs.
Lastly, consider this a commitment to shipping more. At least weekly. Maybe the future of human cognition is partly about untying creative knots.