Ideas Will Be King
In a world where implementation is commoditized, ideas will be king. That world is just round the corner.
It’s been almost two weeks since I took my head out of the sand to fully appreciate the advancements made in Artificial Intelligence over the last few years, and especially in the last few months. The decision was timely, because the last two weeks have been overwhelming.
I needn’t re-state the obvious in this post, so I will focus on my thoughts about how I figure all of this impacts us. This is the perspective of a software engineer — a knowledge worker not that different from a plumber — , a hobbyist creative writer and composer, and general tinkerer with things.
A New Paradigm?
My argument for how we’re going to live with our new super intelligent tools is simple: when we have technology that can trivially implement the kinds of things we take our time to essentially handcraft, our “work” will largely be to direct those tools to make the things we want.
What AI has taken to obscene levels in recent weeks is something that can be easily observed by taking a look at the technologies we use right now, how it augments our workflows, and how tedious our work will be without them. Case in point: the autocorrect I’ve used more than a few times while typing this up.
Even in my current workflow as a writer and editor for a small online review site, I find myself sending out prepared texts to be proofread by our volunteers just to make sure grammatical slip ups and factual inaccuracies are not published as authoritative. Higher level editorial tasks that cover tone, organisational voice and such are things I busy myself with. Were I a manager with a team of writers at hand, the task of creating new content will comprise of at least two things:
- Request the content from the team
- Review and request refinements
Introducing an intelligent tool into that workflow could make this level of work possible, without the other human dependencies (yikes!), and it could scale the rate at which material is created and published.
My use of AI in my writing workflow today goes beyond spell checkers. Whenever I publish on Medium, I get it to read me back my words, and there’s always something I catch when I listen, that I would have missed had I simply read through my work. I’m not sure how many folks will do me that favour whenever I want it, however often I think I need it.
Note two things: first, this is not meant to trivialize editorial work. I’ve been in those shoes long enough to know it’s its own grind. Second, this is not a value judgement about the actual task of writing versus editing and directing. I am simply recognizing a reality that, when the tools we use get really, really good at a certain level of work, humans need not compete.
Aside: High and Low Level Work
Before I started animating with Macromedia Flash, I had learned a thing about prestigious Japanese studios that had their most valuable talents draw the “key frames” of an animated scene, and employed small armies of low-wage artists to fill out the in-between frames. It did not dawn on me then that by specifying keyframes in Flash and letting the software fill-in the blanks, I was engaging in the same dynamic.
I did realise then that those studios reified a power balance I was not comfortable with, and as much as I respected their output, I guessed (perhaps wrongly) that being an animator would suck for the most part, and killed my ambitions about that line of work. I also realised that they recognized a sharp distinction between levels of labour that I became more conscious of with age and experience.
(Fun aside: the world of computing is explicit about levels of abstraction that can — with some squinting and fitting — map on to this distinction. Programmers love to automate these things away when they don’t need to think about them. On the other hand, my colleagues and I back at school were inclined to look with awe at those who worked on low-level code, because they were doing the “serious”, close-to-the-metal stuff!)
Let’s just be honest, okay, there is some value judgement in this distinction, and on average many more people will prefer one to the other. A more down-to-earth view recognizes the nuances of human preference, and allows that in the face of the obvious societal bias in favour of high level work, there is a lot to love about the boring stuff. Some people still swear by manual-transmission cars, for what its worth.
A Glimpse of What Might Come
Yesterday, OpenAI announced ChatGPT Plugins, a way to give their already supercharged AI even more superpowers and access to the outside world. The announcement site included a tutorial on creating a new plugin, during which a GPT was used to write the actual plugin code while the narrator went on to explain the parts we cared most about: what makes a plugin a plugin.
That’s the point!
The goal of the video was not to teach viewers how to write a To-Do webservice with FastAPI. In a different world (that was a few months ago!), he would have copy-pasted the relevant API code into his main.py file. However, the other point of the video was (I speculate) to show us a new paradigm of application creation. And so he delegated that unimportant bit to his GPT and carried on with what mattered more.
In the real world, he would have asked ChatGPT to write the entire plugin.
He made a bit of a show of auditing the GPT’s contributed code; that’s a best practice OpenAI will want to promote explicitly, which many people will flout out of convenience. But with that, he demonstrated again the new paradigm that’s only possible when the grunt work is commoditized and scaled by computers.
Brace Yourselves
It is possible that, if content creation becomes less laborious and less costly, two things will happen:
- There will be more ambitious projects, because costs are low
- There will be A LOT of content, because the barrier is low
Case in point: I’m a storyteller with a lot of things I would love to realise somehow. For people in this situation, there are a number of options at their disposal: writing and publishing, dramatising as a live audio or video production, animating, game-making, etc. Today, these option’s are constrained by the creator’s talent set, social network and access to capital.
In a world where the grunt work involved in realising a story in these forms is feasibly automated, the would-be artists could, if they wished, employ intelligent tools to do the heavy-lifting, intervening when they have specific opinions on how some closer-to-the-metal activity is actually done. Those with enough love for the craft will still get their hands dirty for the love of it, but not because they have no choice.
I’ve been there before. We all have. Sometimes you just need a draft to build upon, or a scaffold to start with. Other times, you’d rather lay it all out yourself, word after word. A proliferation of super intelligent tools can make this choice possible.
This Might Work For Us
If AI is mainstreamed in industry, a lot will change. It could easily be for the better. A lot of low-level jobs might be impacted negatively. It might force such workers to take on higher-level roles and oversee production of whatever they have expertise in by non-autonomous AIs.
It might lead to more nuanced products, services and content, because production will be cheap. It might lead to even more ambitious projects for the very same reason.
This rapid revolution in computing might also be an absolute disaster for the global workforce. The risks of third-order effects are high with something this incendiary, and our world today is in a bit of a mess, if you’ve noticed.
In the meantime, tech adoption is uneven, and there’s an estimated 3 billion people not connected to the Internet (revealed to me by a GPT 🤷🏾♂️). There’s a lot of uncharted territory for humans to explore, and an increase in productivity might unlock a lot of that trapped potential. It’s just too early to tell, and the accelerationists aren’t stopping now. Nor will they be, because progress is a juggernaut that is only halted at greater expense.
It’s best to get on with the program.
If you enjoyed this, let me know.
