The Workforce of the Future

Vivienne Ming discusses our future workforce.
Vivienne speaking at TEDMED 2015.

“How can we make students robot-proof?”

This was the question posed to me a few years ago at the Department of Education. They were designing a 6-week job retraining program, but couldn’t divine which skills to teach that would still be employable 10 years from now. They were hoping I could help them devise a short program to teach some crucial knowledge or skill that we knew wouldn’t be displaced by artificial intelligence in the future. Here is the simple truth I shared with them: there is no skill or knowledge which is robot-proof.

Automation, Artificial Intelligence, and Why Automation Is Different This Time

The concerns of the Dept. of Ed. are well-founded. Quite apart from the hysteria surrounding general artificial intelligence – the existential threat posed by a fully aware computer that surpasses us in intelligence – they and many others are focused on a much more mundane problem: basic AI and other computerized automation will displace the world’s labor force, leaving few jobs for humans. The three largest employers in the world–agriculture, transportation, and natural resources–are all seeing huge advances in robotic automation. For example, precision farming involving drones, robotic weeders, and AI-driven irrigation produces more food but needs few humans. The productivity gain truly is a fundamental good, but what happens to the one billion agricultural workers worldwide?

Importantly, AI isn’t only replacing physical labor, but cognitive labor, and it’s doing so at an increasing rate. We now see automated systems outperforming or displacing humans in medical diagnostics, journalism, financial advising, and a vast array of other industries. A recent paper described a deep neural network that can read the technical specifications for software companies’ APIs (the rules that allow one computer system to interact with another), and can then automatically spit out a simple set of instructions for writing code using the API. It would be a rather minor additional step to have a basic AI write the code itself. Despite all the focus in recent years in teaching students to code, it seems unlikely to me that simple programming will be a viable skill 5 or 10 years from now in the same way it is today. I’ll simply “hire” an AI contractor, giving it a set of specifications and even having a conversation around the details. It will quickly and easily spit out prototypes and update them based on real-time feedback. While this sounds wonderful to me as an entrepreneur, software developers might feel like they bought a home just as the real estate bubble burst.

AI Workforce vs. Human Workforce?Techno-utopians may claim that AIs will free everyone to be artists and doctors. They imagine themselves freed of the burdens of rent and the need to take a job just for the paycheck, spending lives of purpose solving deep problems. Our schools and other social institutions, however, are simply not designed to produce a workforce full of problem-solvers. It is much more likely that we would have a world in which the labor of some is worth more than an AI, but the labor of the vast majority is worth less. What a profound divide that would be.

Craftsmen and Their Tools

Research has shown, rather dramatically, that knowledge and skills, and the grades, test scores, and degrees associated with them, are simply not predictive of employability and other life outcomes. Yet schools and so many job-training programs focus exactly on these: how to program, how to factor a polynomial, how to write a grammatically correct sentence, or how to sketch the human form. They are valuable skills, but only in the hands of someone empowered to make use of them. These are just the tools that craftsmen employ, not the craft itself. What predicts life outcomes is the quality of the craftsman. A large and growing body of research links success with qualities like general cognitive ability, metacognition, mindset, emotion regulation, and creativity. These are attributes which we have described as meta-learning–the deeper abilities that enable learning.

A further, fundamental problem is that no tool is robot-proof. There is no basic skill or knowledge which we cannot eventually build an AI to perform more economically than a human. Tools neither differentiate people from one another, nor protect them from robots. Instead of trying to guess which skills kids need to know 20 years from now, we should build craftsmen who can master any tool. A craftsman without their tools is hobbled, but tools without a craftsman are entirely pointless. To robot-proof our kids, we must develop their meta-learning skills, producing a generation of problem-solvers.

A human story

With AI’s providing all of the tools, the future of work is the hyperinflation of work: you’ll show up in the morning, and it will be a different job by the end of the day. The only job description in the future will be that of a problem-solver, with every day posing a different problem (and it sounds damned exhausting). But imagine what a society full of such craftsmen could accomplish with a toolkit full of AI tools. What could be accomplished if we truly were a society of problem-solvers, of craftsmen?

I am a huge advocate of the potential of machine learning, AI, and even the eventual power of augmented intelligence and neuroprosthetics. They are a foundational part of a world in which I want to live. But this is fundamentally a human story, not a technological one. No one is going to stop the rise of AIs. We need to match it with the rise in social institutions built on the core principle that everyone can be amazing. But it takes years, even decades, to “build” an amazing person. If technology continues to outpace culture, the results will be catastrophic.

We don’t have to accept that outcome. While there’s no online course or six-week job retraining program for meta-learning, we know how to develop it over time. We know how to build into kids a belief that their hard work will pay off. The irony is that the solution to humanity’s place in a futuristic world of robots and AIs is as old as it gets. The things that will make us robot-proof are the very same things that are predictive of life outcomes of both kids and adults today, and have probably always have been throughout the history of humankind. The best way to robot-proof your kids is to make them all the more uniquely human.

This guest blog post is by TEDMED 2015 speaker Vivienne Ming. You can watch Vivienne’s TEDMED talk here.