The Automation of Mediocrity

AI is the opportunity to level up as humans.

We’re in a specific cultural moment — Artificial Intelligence breaking into the shared consciousness of the so-called “knowledge industry”. If you are reading this, you’re likely to be in that bubble. This moment has been triggered by the release of ChatGPT, first of many general purpose “knowledge” AIs to be released this year.

Your social feed might be full of angst about AI, so to add to that would be boring. My interest is in how AI interacts with humans and how it can help us flourish and level-up our creativity. What follows is a highly opinionated and speculative — but original — journey into the near future (i.e. a rant).

The Deck Chairs of “Knowledge Work”

The AI hype cycle is both feeding and being nourished by the anxiety of the average “knowledge worker” — the anxiety of being (fully or partially) automated. Industries like consulting, law, design, coding etc. have never known existential dread. As white collar professionals, we may have intellectually understood the threat to blue-collar work, but we never had to live it and feel it ourselves.

But now we’re at the cusp of a seismic shift in which AI will automate knowledge work — and a lots of things are going to change, very fast.

ChatGPT is a prolific writer and a fast and broad researcher. It can write articles and essays. It can write code. It can generate ideas and suggestions and recommendations based on prompts. And it improves. Over time, AI will out-compete in anything that is purely the organization of knowledge or pattern-matching and pattern-applying of knowledge.

If that’s an accurate description of your “knowledge work”, your job will be in jeopardy. If I can generate my own Terms of Service document, I won’t have to pay my lawyer thousands for it (who uses a template anyway). Twenty years ago I was paid to write SQL queries, but perhaps nevermore. For a minor health issue e.g. a skin rash, almost anyone might prefer an AI doctor for a data-driven diagnosis. And so on.

Data-Driven = Mediocre

But the thing is ChatGPT writes mediocre essays. It’s great at finding information that already exists and putting it together, but there is something …. boring about it. Try reading ChatGPT essays yourself and see. It never really surprises. It doesn’t make intuitive leaps or insert personal anecdotes or build narrative tension or make creative juxtapositions or synthesize disparate fields of knowledge …. &c

[Or: an AI won’t parenthesize, go off on an unrelated tangent, apologize for it (sorry) — then resolve back to the original theme to make a point.]

The italicized words above are part of the domain of creative thinking. Regular thinking uses known data within the rules of logic and stays within the existing possibility space. AI can do this. Creative thinking moves into adjacent possibility spaces and that movement has nothing to do with data or existing knowledge. AI can’t do this but humans can.

AI maximalists say — just wait, AI will improve and eventually do all of that. But they’re wrong because the Rubicon between thinking and creative thinking is un-crossable for machines. AI trains on data, AI needs data, AI is ‘smart data’. But the leap into creative thinking is qualitatively different, not just an increase in degree, because it’s not about data. It’s water to air; or maybe better air to no-air, like humans leaving our own atmosphere.

[Side note: Astronauts can leave Earth but only by taking their home atmosphere with them; and in the same way, an AI can take data with it and use randomness to simulate certain creative skills, like juxtaposing thoughts randomly — sometimes it generate jokes, but it doesn’t know which one to tell or how to tell them (no data!).]

So let’s (re)define mediocrity as this — any purely data-driven, logical task or process undertaken by a human. These are “mediocre” knowledge jobs.

AI Eats Mediocrity

Soooo much of our knowledge work falls within this “mediocre” catchment area, because industries rely so much on the arbitrage of access to pattern-matching and pattern-applying knowledge. Take law, for instance. Lawyers know more than you about law or at least they can access the right knowledge quickly; they can then reason with that knowledge and create outputs much more quickly than you. Good lawyers may think creatively as defined above — but most are logical and mediocre. AI can access all the knowledge 10,000x faster and do all of the non-creative “thinking” for you.

[Picking on lawyers as an easy example. Applicable elsewhere too.]

The genius of the AI revolution at this tipping-point is that it being made by creative-thinking humans who are using knowledge from the non-creative realm to build machines … that can do non-creative knowledge tasks to replace humans. This feedback is possible due to 30+ years of indexing of knowledge on the Internet. And it means that the development of AI and its cannibalizing of knowledge work will be fast.

So now the good news.

Our gut-feeling is that we can train humans to evolve quickly and open up enormous new possibility spaces on the right side of the diagram (creative thinking). Whole new industries and categories will emerge. It happens at every technology-driven step function in history, and it will happen again.

But it will disrupt a lot of the status quo — both up- and down-stream from knowledge work.

Time to Level-Up

A foundational myth of our civilization is that to do something well you have to be able to express HOW you do it. This is false — as Taleb says, birds don’t know aerodynamics. For knowledge work on the left side of the diagram, you can transmit the knowledge of how to do the work —a process. But for the right-side creative thinking skills, it’s very difficult to even talk about them meaningfully, let along break them down into process.

So if you consider learning as assimilating knowledge as process, you can’t ‘learn’ creative thinking. But you can train by doing. For the non-mediocre skills that only humans are capable of you must train by doing. No process can tell you how to make a joke, but you can get up and do it, get feedback, and do it again. Iterative learning replaces process learning.

The level-up will play out over the next decade plus, moving our knowledge work to the right and replacing mediocre knowledge workers on the left. It’ll be messy and will come in fits and starts, with reversals and chaos and “unfairness” and winners and losers, as always.

And it will also be a level-up of the way we learn, creating enormous down-stream disruption in the education industry. The kids are being educated in the wrong way for the wrong jobs, and they know it even if they can’t always articulate it. When we put them into an arena where they learn by doing their creative skills, they shine. Then we add AI bots to the mix, and they respond by upping their own creative “performance”.

It’s fascinating stuff, but at the beginning of such a wave of disruption there’s never much science to it. The only way to work it out is to run many many experiments and iterate as fast as you can. As in any revolution — practice and technology first, theory and science later.

Evolution happens in bursts. There is nothing more exciting than being right at the cusp of such a burst of change, when you can literally feel the undercurrents shifting. If you’re early, you can be a part of the change, try to make the change human-hearted and actions have outsize influence. I’m convinced that we are entering a Creative Age, where the birth right of every human — a unique creative mind — can be expressed on a scale never seen before in history.

Footnote: our minds created the machines that are now automating part of what our minds do, which now shows us clearly what is uniquely human about own minds. I think that’s a beautiful piece of irony.