All we need to do is make sure that we’re in a position to benefit from uncertainty and volatility instead of being harmed by it. ~ Nassim Taleb

If you’re confused about the impacts of AI, you’re not the only one. Some predict that AI will destroy civilization as we know it; some predict that AI will revolutionize our lives. Some say that our jobs will be gone; others that we’ll be able to do those jobs better, and there will be lots of new jobs we haven’t even conceived of yet. The one thing each of these seems to have in common is the impact that AI is currently having on the speaker, i.e., the impacts are seen only through the eyes of the beholder. Certainly the impacts of AI are both highly uncertain and highly personal.
As both Niels Bohr and Yogi Berra have said (in almost identical words!), prediction is difficult, especially about the future. So I won’t try to predict AI’s ultimate impacts. Undoubtedly they will be profound, and compounded of good and bad. But I will provide some facts to help think about AI’s impacts, and how to adapt to them. For one thing is certain: the more adaptable we are, the better our chances of avoiding being devoured by the AI Monster.
Impacts on individuals
As a few of you know, I was involved in some of the early attempts to deploy AI in the real world, in the 1980s. At that time, the practical emphasis was on development of “expert systems.” These required gathering the knowledge and experience of experts in a given field, and converting it to a set of rules along the lines of “If some condition, then do this.” And the results were sometimes impressive. For example, a psychological diagnostic system was developed that was as good at dealing with psychological trauma as human therapists.
A lot of the current effort in AI is still this: replacing the “man-in-the-loop” for foreseen situations with a set of rules. The major difference is that our increased computing power and the advent of Big Data are allowing developers to vastly increase the scope of what is “foreseen” and our ability to act swiftly. As a result, we have self-driving cars with accident rates below those of human drivers. We have “apps” that write computer code more rapidly than human programmers, and (if not now, eventually) with fewer errors. AI has growing capability to diagnose diseases, to dig through case law to find relevant precedents, even to point to promising areas for research.
This is due to the evolution of AI from task-based – if-then rules based on a defined body of knowledge trained to act; to generative – systems able to change themselves in response to new knowledge. This is often called “machine learning.” App-adaptation might be a more apt term.
This makes it seem like the doom-and-gloomers are right. The AI Monster will devour us all. Well, not quite; it’s complicated. And if you’re going to escape from the AI Monster [maybe] lurking in your closet, you need to look at your situation through several different lenses.
• Current employment. The most quoted unemployment rate in the US is currently at 4.3%. Personally, I find labor force participation more meaningful; for 25-54 year-olds it’s about 84% (pretty high). Conversely, it is in the 60% range for those 16-24. This implies that level of experience is important. If young people aren’t being hired to do a job, or, worse, those with experience are being laid off, that’s a sign that the job is in jeopardy.
• The job itself. It’s almost become a truism that AI won’t displace those in the “high touch” professions (e.g., family services, sales) for a very long time. But even there, many of the more routine tasks in health care (e.g., reading x-rays, billing) are being taken over by AI. On the other hand, many knowledge workers are at extreme risk. It seems like every week thousands are being laid off by one of the Big Tech companies. Whatever the field, jobs requiring flexibility, especially those developing policy or strategy, are likely safe for a very long-time.
• Cost-benefit. Generative AI systems are expensive to develop. They require massive amounts of domain-specific data, tremendous computing power, and extensive testing. And lots of electricity and air conditioning as well! The greater the complexity of a given job, the less attractive it becomes for AI – the costs may outweigh the value of displacement.
I asked both Claude and ChatGPT (two of the leading AI apps) “What will be the impact of generative AI on future jobs in the US?” Both gave lengthy responses. Claude’s was better written, and I found more generally useful:
The pattern with previous automation was that it tended to displace routine physical work — assembly lines, data entry, basic bookkeeping. Generative AI breaks from that pattern in an important way: it’s targeting cognitive and creative work first. Paralegals, junior analysts, radiologists reading scans, software developers doing boilerplate code, copywriters, customer service agents, financial advisors doing routine portfolio management — these are all in the crosshairs in the near term.
Jobs with the highest short-term resilience tend to share one or more of these traits: they require physical presence and dexterity in unpredictable environments (plumbers, electricians, nurses doing hands-on care), they depend on genuine human relationship and trust (therapists, certain kinds of sales, clergy), or they require highly contextual judgment in novel situations (senior executives, certain researchers, emergency responders). …
The most likely scenario is not a clean “AI takes jobs, new jobs appear, equilibrium restores” story. What’s more probable is a bifurcation: a class of highly-educated, adaptable workers who use AI to become extraordinarily productive and well-compensated, and a much larger group of workers displaced faster than they can retrain, in a demographic environment that offers fewer natural correctives than past transitions. The demographic plateau softens the raw unemployment numbers, but doesn’t address the human cost to specific workers and communities.
ChatGPT, on the other hand, gave a similar answer with two interesting twists. First it provided more detailed impacts for important sectors (e.g., tech and IT, government, education, manufacturing). ChatGPT also looked at the impacts of AI over time. In the short-term, it predicted job losses, some increases in joblessness and reduced hiring of new college graduates. In the medium-term (3-10 years), ChatGPT saw restructuring of existing jobs (skill in using AI becoming embedded in many jobs) and the rise of new categories of jobs (e.g., “prompt engineering,” data governance, AI trainers and testers). In the long-term (10+ years) ChatGPT weaseled out of prediction and instead fuzzily pointed toward a bifurcated future.
From all of this, let me offer the following guidance to all of you worried about becoming a victim of the AI Monster.
- No matter your profession or vulnerability, become familiar with AI tools. AI is here to stay. And it’s getting better at doing routine and non-routine tasks. If you can use it, it will increase your value to your employer/customers/clients. More importantly, using AI tools can actually increase your job satisfaction.
- Work to redefine your current job to require more non-routine problem-solving. This will help you to further develop AI-complementary skills: learning, planning, creativity and decision-making.
- Continually increase your knowledge about how your job/profession plugs into everything around you. Most importantly, focus on your own personal value proposition. But be brutally honest with yourself: if you’re mired in the routine, your value-added may not be high, i.e., you may be vulnerable.
- Increase your value. Certainly, taking on more responsibility is one obvious way. Perhaps more importantly, increase the complexity of what you do, especially if your work is largely repetitive or routine. If you can’t, best brush up your resume, and your skill set (see #1). If you can, find a way to apply your skills in a “high touch” environment.
- An important way to do both #2 and #3 is to expand your own personal network. Try to determine how people with your skills are used elsewhere. Perhaps more importantly, become a linchpin: a conduit for useful information. Involvement in professional societies can help you do this.
- Develop a “Plan B” for yourself. Identify where your current skill set might be of use. Continue to expand your skill set (again, see #1).
Impacts on communities
Claude’s response quoted above points out that AI will impact both individuals and communities. In my next post, I’ll explore potential impacts, and suggest things communities can do to better buffer themselves against potential harm.

