The Best Resistance is a Demonstrated Alternative
It's not inevitable unless we make it so
Thank you for being here. Please consider supporting my work by sharing my writing with a friend or taking out a paid subscription.
There’s a lot of talk these days about AI and inevitability. Every big tech CEO (read: person with a vested financial interest in a certain outcome) seems to be constantly talking about the way that AI is going to completely reshape the economy by automating cognition or at least automating the products of cognitive labor.
While I’m sure many of you reading this are in the camp that views such claims with deep skepticism, I would like to suggest that this is not all fairy floss. There is something to what they are saying.
While the macroeconomic disruption may not yet match the microeconomic reality of what these tools offer, the disruption will certainly continue to happen, even if the tools get no better than they currently are today. Even pre-genAI, it doesn’t take a particularly eagle-eyed analysis to see that every corporation is bloated, with many of the tasks that make up full time roles ripe to be replaced by a machine.
Inevitability comes in two flavors
To understand this, you can think of two flavors of inevitability: strong and weak. Strong inevitability argues that a certain outcome is set, and it is only a matter of time before it becomes reality. This argument assumes that there is nothing that can be done but wait for the series of events to play out; that the outcome is predetermined by the current state. This is the kind of inevitability that seems to drive much of the rhetoric out of Silicon Valley these days. It is also the kind of reasoning that undergirds the predictions of dystopian outcomes like in If Anyone Builds It, Everyone Dies.
But most of the time, what is labeled as strong inevitability is really weak inevitability in disguise. Weak inevitability is the future that is highly likely to occur so long as the dominant forces in the current system remain intact. I would argue that many of the predictions being made about AI actually fall into this category. They are inevitable if and only if the structures and systems that are currently in place remain so. The only way to break the inevitability is to change the broader system.
Once you shift from strong to weak vulnerability, you move your focus from the technology itself to the systems and structures around it. This is the challenge before us. Generative AI is here, and whether you like it or not, there is no going back. That’s just not how technological progress works. However, just because we can’t “uninvent” generative AI doesn’t mean that the future is pre-determined. It just means that the future they are predicting is inevitable to the degree that we don’t do anything to respond.
A view from my seat
One of the unique opportunities of being on sabbatical this year is that I’ve had a lot of margin to attempt to get close to the cutting edge of these tools. Most days, I spend several hours in Claude Code (via Conductor) building software. Outside of writing software, I am almost constantly testing AI tools across a wide variety of the other tasks that make up my days—everything from searching for information, brainstorming new ideas, writing budgets, to analyzing data.
All this to say, I am witnessing first hand the kind of microeconomic impact that AI is having. The Generative AI tools we have today are absolutely good enough to help with a lot of my work. Many of the places I’m using AI are bringing me real value.
Just this last week, I used Claude to help develop a budget proposal. It was a very helpful tool to ensure that my proposed budget was meeting the budget guidelines and helping to reformat and draft budget justifications. Of course, I had to review and make some minor corrections, but it saved me a lot of time, time which I was able to spend thinking more deeply about the actual plan for the project. To write an effective proposal, you still need to have a good idea, but so much of what goes into writing a proposal itself (and especially the nitty gritty details of the budget) are things where automation should be readily welcomed. The budgets themselves are important things, but the process of drafting and formatting them lies squarely in the quadrant of tasks that we should outsource to AI.
In this vein, I think of generative AI as a more capable version of Excel. You could make budgets before the spreadsheet existed, but it was a lot more challenging and required a lot of cognitive effort that could be spent more productively elsewhere. I understand that the generality of generative AI means that it is not an apples-to-apples comparison with the spreadsheet. And yet, if we consider applying generative AI in narrow application areas, the comparison might be more legitimate than you think.
The point here is that AI is disrupting our world and will continue to disrupt our world. It is truly useful in a variety of domains, and even if the model progress freezes (and it won’t), there is a certain level of disruption that is, if not inevitable, pretty darn close to it.
How to dismantle strong inevitability
But the silver lining is that most of what is argued to fall in the category of strong inevitability does not truly belong there. If most strong inevitability is actually weak inevitability in disguise, then what should we do?
Well, the first thing is to make this clear. We should be thinking critically and surgically about the weaknesses in the arguments that presume strong inevitability. We should be building a case to go from the vague statements about pick-your-percentage of white-collar work being eliminated by AI, and actually dig in and make a model to explain how such a drastic shift might occur and to highlight the assumptions that must remain true for that disruption to occur. The train is moving quickly, but the rails have not yet been fully laid. The macroeconomic disruption is still slower than most people appreciate. There is still time, although perhaps not as much as we would like, to take stock of things and change course.
But more importantly, we must act. There is a certain flavor of “activism” today that feels focused on what we say and what we ask other people to do. Of course, this is a vitally important part of creating any change, but it is insufficient. What is even more important is the ability to pivot from sharing ideas that articulate the way we ought not to go to creating a vision that says “follow me” in the way we should go. The best resistance is a demonstrated alternative.
This feels to me like one of the only ways to effectively push back on the inevitability rhetoric. We can’t just rail against the genAI hype without recognizing and acknowledging that there is indeed something to these tools. They are not all that their creators make them out to be, but they are useful and effective in many contexts.
We need to do our best to stay up to speed with what the latest tools can do. Not so that we can mindlessly integrate them into our work, but rather so that we can make sure that our theses for alternative directions remain sharp and relevant. We should reject strong inevitability hypotheses, but not be so quick to dismiss the potential for weak inevitability. In many cases, if nothing changes and society does not adapt, the kind of things that these CEOs are saying will come true.
The real challenge in front of us is to help society adapt to the coming changes. And that happens one person at a time, not just by saying something, but by doing something and then by getting someone else to follow in your footsteps. Speak up and then step out.
Don’t just push forward a proposal for someone else to do something about it, do something about it yourself. Build an alternative. Make your voice heard, but more importantly, build an alternative way.
Got a thought? Leave a comment below.
Reading Recommendations
Stumbled on this 2021 piece from Shannon Vallor in Noēma titled “The Thoughts The Civilized Keep”. Interesting to see how, despite all the technical progress, we still have a long way to go until we address the most significant problems posed by AI.
In an era where the sense-making labor of understanding is supplanted as a measure of human intelligence by the ability to create an app that reinvents another thing that already exists — where we act more like GPT-3 every day — it isn’t a surprise that GPT-3 might be mistaken for the AI breakthrough that will spawn true machine intelligence. But as AI researcher Gary Marcus and many others have acknowledged, that goal awaits in a different direction. If machines ever do join us in the domain of understanding — if they become able to think, know and build new worlds with us or with one another — then GPT-3 will be a footnote in their story.
But even as someone who thinks about AI for a living, I don’t find myself worrying much about when, or if, machines will get there. I find myself worrying about whether, by the time they do, we will still be capable of thinking and understanding alongside them. We can get by and endure, for a while, by riding the coattails of those who labored before us. But not for much longer, unless we repair and restart the engines of thinking for future generations — the cultural institutions, social practices, norms and virtues that valorize and enable, rather than penalize and suppress, the shared human labor of understanding.
Humanity has reached a stage of civilization in which we can build space stations, decode our genes, split or fuse atoms and speak nearly instantaneously with others around the globe. Our powers to create and distribute vaccines against deadly pandemics, to build sustainable systems of agriculture, to develop cleaner forms of energy, to avert needless wars, to maintain the rule of law and justice and to secure universal human rights — these are the keys to our future.
Yet they are all legacies of past labors of understanding that even now we wield with increasingly unsteady and unthinking hands. Of course, these achievements would all be impossible if Whitehead’s words were not in large part true. But we have failed to seriously ask the question that should have followed: “What thoughts do the civilized keep?”
Appreciated this post from Steven Mintz on AI an labor.
[T]he case against degrading labor cannot rest on efficiency. It must rest on human dignity. And arguments about dignity require precisely the moral vocabulary the labor question once sustained and that we have largely abandoned.
The Book Nook
Slow progress lately in the midst of a busy season, but still doing my best to make my way through Kingsnorth’s latest.
The Professor Is In
I’m excited to see my teaching schedule for the fall come into focus. I’ll be back teaching my embedded systems class (E155)—this time co-teaching with one of my colleagues, which is always a treat—and also co-teaching a new class tentatively titled AI for Engineering with a different colleague in the department. I’m excited to get a chance to dig into more of what these tools mean for engineering practice and how we should be helping students to grapple with them. More to come!
Leisure Line
Continuing to spend a few hours each week in the outfield coaching #1’s Little League team. Fun to see the kids getting better and better with practice.
Still Life
Found this guy hanging out in the front yard last week. Your regular notice that y’all are sleeping on how amazing the iPhone macro mode is.
I hope it's ok if I press a bit here on a key point that weighs on questions of inevitability, strong and weak.
Can you give a more specific description of the deeper cognition you were able to do about the grant once you were freed from the distracting detail work like compiling the budget?
These are kind of claims that I see from some people who have been making use of generative AI and I so far have not had that experience. Perhaps for two reasons--the first that is for me the detail work often illuminates the "deep thinking" work. Working with a group a while back on a grant application, I felt that the question of what we needed money for and exactly how we would try to spend it was a substantive problem. Writing the top-level narrative wasn't all that hard--it was a set of ideas that we were all comfortable with and committed to. Figuring out how the grant would let us enact those ideas in new ways was where all the "deep thinking" had to happen. So I feel as if automating would have given us a budget that looked "typical" when in fact we needed to think atypically about it. The second reason is that if the "deep thinking" is hard, it's not because I don't have the time, it's because there's something in there that I don't know how to think about, that I'm uncertain about, and a lack of time isn't producing that ambivalence.
But maybe this is just me in my domains, where I'm not sure what I'm meant to be freed from or what I might be liberated into. What specifically did you feel you were able to think about better once the cognitive load of preparing a budget was off your plate?