I've thought for a long time that "efficiency" is often over optimized for. The seed was probably set by hearing Alan Watts playfully ask if dancing or love making should be efficient. Efficient music? Just one note! he teased.
I recently attended my uncle's funeral. He was a man who achieved much in his life professionally and personally but his legacy seems mostly to be that which he did for others. A teacher for over 40 years he inspired countless students to be better writers and he also spent much of his life (successfully) championing the art work of a good friend who succumbed to HIV in the 80s.
As I ruminated on his life and reflected on my own I recognized my own myopia. My concerns about my career, planning for the future and building "something cool", are, I realized, something short a "life well-lived". This led me to thinking about my work with AI.
During a conversation with an engineer friend, I suggested that AI was best thought of as a "force multiplier". He suggested that all (useful) technology can be described this way. There is a lot of heat at the minute about AI's potential for good and harm. The most recent polling seems to suggest this conflict is not between one group or another, except at the outer fringes, but rather inside most of us, especially those of us who use the tech.
With the launch of ChatGPT I was all-in. So much of the agency work I was doing could be augmented and improved with the use of an agent. I'm a one-man shop, who often starts with a conversation or two with a client and encompasses the entire process to shipped product, implementation and sometimes promotion. PM, front-end, back-end, devops, QC, and even graphic design, copy writing, and marketing. I am the classic jack-of-all-trades, and being able to workshop site themes, or troubleshoot a technical issue with another, even one that made mistakes immediately meant that I could not only do things faster but meant that I moved up to a more supervisory role and aspects of a project that just wouldn't be in scope due to time (read: money) were suddenly in scope and shipped. My work got better because instead of grinding away in low level tasks or yak-shaving some annoying task I could look up and see if the project was moving toward the thing we were building for.
For this reason I am often bemused by the argument that AI "is just a toy" and "can't really do anything useful".
I have shipped enormous amounts with AI. It has been a partner, a force multiplier for my work. Not a replacement. GIGO is still an axiom. The spark is the human idea. LLMs live inside the borders of their training. We do not. We have ideas, thoughts and inspiration. LLMs are execution agents. They will do what we ask, quickly and often efficiently, so we should not underestimate the very challenging tasks in front of us in regard to alignment, safety, and reliability. Poorly aligned they will eagerly wreak havoc, then follow up and ask "what's next, boss?" We also need to ask the right questions and charge these execution agents with noble tasks.
I will suggest we demote efficiency as a first step. We've gained massive leverage there already. Between computers, the internet and agents, we're already orders of magnitude more effective than people were just 40 years ago.
Human flourishing, on the other hand, is not often put forth as an optimization target outside some of the more poetic missives coming out of the marketing department. But what about bringing this into engineering and actual organizational direction?
There's a lot of very valid concerns about job loss. There were also very valid concerns when ATMs were rolled out, even among bank managers. Why pay a human teller when a machine can do the job 24/7, never takes a vacation and will almost certainly have a lower error rate?
In fact the number of tellers increased rather than decreased because ATMs made locations that were hitherto unattractive due to lower demand viable, and in fact teller jobs actually grew slightly faster than the overall job market.
Will this happen in the current era? Hard to say but certainly technology that makes more work possible has not yet resulted in fewer human hours.
A secretary of the bygone era could only produce so many documents, so many letters typed. Word and mail merge means that even someone with my pitifully slow and error prone typing can produce hundreds or even thousands of typo-free letters inside an hour. That tech is 30 years old. Did companies start putting everyone on part-time? Did they massively shrink staff because they now could produce the same with fewer staff? No. They increased the volume of work. In economics this phenomenon is called Jevon's Paradox.
Did the "secretary" as a job get phased out? Yup. But they were replaced with an administrative assistant that is arguably a much more interesting and varied job and almost certainly one with more opportunity for advancement and yes, room to flourish. And administrative assistants are more broadly useful which creates an incentive to have more of them.
Staying on the theme of human flourishing it's valuable to ask: how can we optimize for it? What does human flourishing even mean?
Maybe it's agency or dignity. Perhaps craft or service or learning. For others health, social trust, or meaningful work. And others still it is the relentless pursuit of beauty.
How can we use these new tools to improve the lives of the humans around us? Current roles are being displaced. New roles are being created. How can we align these with what is best about humans? How can we use AI as a force multiplier for meaningful work and outcomes that improve lives? What are we trying to do? Perhaps we need to ask "squishier" questions.
Why do we dance? Make love? Why do we help others? Create or seek out art? Have families? Make friends? Laugh? Play?
These are not acts of efficiency optimization, they are something else. The things worth doing for their own sakes and the reason why we do the other things like go to work and try to make that more efficient - so we can have more of the good stuff. The stuff of human flourishing.
Now that we have the most powerful force multiplier in history it's time we ask: what are we multiplying toward?