The Can-versus-Should Question

I co-presented a workshop at SXSW last week on system dynamics, helping people get started, grasp the basics, and reach a point where they can follow up and start putting system dynamics into practice in their own work and life. 

True to SXSW style, both at the workshop and out-and-about, I also had the privilege and serendipity to meet interesting humans from a wide swath of backgrounds: engineers, musicians, managers, health advocates, professors, artists, strategists. SXSW is probably the best cross-disciplinary conference I’ve encountered. 

One anecdotal difference from the previous times I attended in 2015 and 2016: the 2024 vibe felt (to me) less idealistic about technology and more focused on benefits versus drawbacks. This is good, in my view, if slightly less fun: a reasonable trade-off for a maturing technology industry. 

And these are the kinds of discussions we need to have more of at all levels of society: what do we want from our technology? What don’t we want? Which capabilities are likely to be net beneficial and which are likely to be net harmful, and how might that change over time? Which capabilities should have been rolled out yesterday, and which ones aren’t ready for rollout yet or anytime soon? 

Learning from experience

We let the internet explosion just happen to us, more or less, which was at first awesome and later terrible as largely unregulated services harmed teens’ mental health and eroded privacy and consumer rights. The more cautious conversations occurring now are perhaps a sign of lessons learned from that experience. We do seem to be doing better so far with AI, where stakes are even higher, since conversations are happening sooner about the benefits and risks.

Although jobs are far from the only concern as AI advances, I find myself thinking about them as a near- to medium-term change likely to affect many people, myself included. And as jobs begin to be replaced, the can-versus-should conversation will likely grow in importance. (As a counterpoint, Noahpinion published an interesting essay this morning on why he thinks AI may not eventually replace most jobs, and/or the jobs humans do might not change too much in the future. I don’t think I agree, but the essay is worthwhile; after reading it, I’m interested in looking deeper into trends in compute costs over time, the hard constraints on the amount of energy we can produce with current technology—since energy production releases heat into our atmosphere and we can’t do an infinite amount of that—and how future energy-efficient technologies might change things by making compute/AI cheaper and more abundant).

For now, here are a few categories of jobs across the knowledge-work and physical-work spheres, facing various amounts of risk as I see it: 

Knowledge-work sphere

Low-skill knowledge work, few barriers to entry: These jobs are already being replaced in large numbers: front-line call-center roles, translation of non-specialized and non-mission-critical documents, data entry. 

High-skill knowledge work, few barriers to entry: These skill-based professions are in danger because of their relatively un-gated nature, which has increased economic mobility and career-switching potential in the past, but I think the party is ending. These jobs may include non-specialized software engineering, user interface design, journalism, copywriting, general consulting and data science, HR, marketing, and back-office operations. 

Knowledge work, gated professions: These industries have historically protected their people through licensing or guild membership requirements for professionals such as doctors, lawyers, librarians, therapists, dietitians, and screenwriters. The exclusive-club approach has also created some shortages that actively harm people, like shortages of primary-care doctors due to artificially constrained numbers of medical residency slots, so there is such a thing as going too far. These gated professional roles are probably among the safest jobs for now, not because they are hardest to replace with AI in theory, but because they are hard to replace in practice based on the current rules of the system. Their moats are constructed rather than intrinsic to the profession. They should and probably will defend those moats while using AI in a way that works for them.

High-skill knowledge work, mission-critical jobs: These jobs will be relatively late to the replacement game because they involve areas where businesses (for now) see a need for human oversight and judgment, despite the costs of humans. They include mission-critical or highly sensitive information security and software and hardware architecture/infrastructure design jobs, AI risk management, C-level executive roles, and entrepreneurship (after all, are you going to fire yourself?). But even these jobs may eventually be replaced or dwindle in number, in my view. I am preparing for a future where I do not work. It’s disconcerting, but I don’t mind thinking about it, because it’s better not to be surprised, or to be positively surprised if there are plentiful roles or better things to do. 

Physical-work sphere

Low-stakes physical jobs: May be replaced soon-ish by robots, but probably over the next several years. Landscaping, farming, fast-food preparation, grocery checkout and stocking.

High-stakes physical jobs: These jobs will probably be okay for a few years, since robots are a few years behind software-based AI as real-world edge cases delay real-world rollouts. My mom immediately grasped it via a football analogy when I explained the difficulty with getting self-driving cars to sufficiently master edge cases: she said something to the effect of, “They’re stuck on goal in the red zone!” Indeed, I expect human providers of in-person services where missteps can be disastrous—driving, nursing, caretaking, childcare, and so on—are all right for now, but eventually they also will face the encroachment of robotic replacements. 

Physical work, gated professions: Like gated knowledge-work professions, some unionized or guild-based jobs are protected more by constructed moats than by intrinsic irreplaceability. The more repetitive or low-stakes these jobs are, the more they will struggle to maintain their moat against corporate pressures. The more high-stakes or high-skilled they are, the better this will go, and the longer they will be able to hold out. As a rough estimate, I think electricians will probably do better than plumbers, who will do better than factory assembly-line workers. 

Physical work, owners and entrepreneurs: As long as they can get customers and make a profit, they can keep going. Think cafe and bakery and restaurant owners, club and bar owners. Humans like to spend time around other humans, and that probably won’t change. The owners will probably employ robot employees eventually, though. Could it be fun to own a restaurant at the end of the universe? Maybe. 

Can versus should

A work mentor of mine used to invoke can versus should as a touchstone. We can do a lot of things, but that isn’t—or shouldn’t be—the only consideration. What should we do with the time we have and the skills we bring to the table and the future we are shaping? What’s the right balance?

And how far can or should the systems we build go in replacing work previously done by human hands and minds? How will energy costs and energy efficiency play into those dynamics over the next several to many years? 

A strong risk culture in an organization adeptly navigates “can versus should” tradeoffs, considering not just short-term profit motives but long-term advantages and disadvantages for the company, its customers, the economy, society, and the planet. 

We need that now more than ever, across companies of all sizes, in various industries, and also at the individual level in our own choices and actions. What’s your can versus should balance? I’d love to hear about it.