AI and Moravec's Survivors
Some will be replaced, others augmented. What about the uniquely human?
Moravec’s Paradox famously revealed many soft skills will remain safe from robots. An AI can rank billions of potential chess moves in a nanosecond. But, once it has humiliated the world’s grandmasters, expecting any version of it to stow all the pieces back in boxes is unrealistic.
So how will AI impact on people earning from soft skills, that it can’t replicate? That depends on what it’s being told to do. Cut labor costs? Or maximize inclusion and opportunity? That in turn rests on who controls the platforms on which AI sits, and the data it can access
This month we rolled out our research with a range of stakeholders focused on what we can reasonably expect AI to do for humans lucky enough to remain economically essential. It is extraordinary how this tech could usher in an era of empowerment, progression and development of human capital.
---
To focus the issues, it’s helpful to take one sector as a microcosm. Consider- as example - the HVAC industry [Heating Ventilation and Air Conditioning]. Simplistically it can be divided into three bands of workers each with a reasonably predictable fate:
Designers: The knowledge workers who draw up plans, specifications and budgets for any building seem ripe for displacement. Once the parameters of any project are known, an AI can generate blueprints better, cheaper, faster than trained drafters.
Site surveyors: Folk who uncover the data required to plan services in a building will be augmented. They will still be needed to instigate measuring, and assessment of conditions like sunlight through a window. But it’s easy to anticipate the handheld devices that will electronically do most of their leg work.
Installers: Anyone who unloads vans, unboxes components, prepares walls and pipes, bolts everything together, assembles the wiring and runs tests in each room is a Moravec Survivor. A robot that could do the work would cost considerably more - once all the risks, transportation and complexities are factored in - than the $30 or so an hour someone with reasonable handyperson skills can command.
For Moravec Survivors, the danger from AI is that it’s being hitched to employee scheduling systems, now ubiquitous for managing interchangeable workers. These breadwinners are vulnerable to AI not because it can do all or parts of what they do. It’s because, given a pool of fungible individuals needing work, it can so ruthlessly figure out how to get each project accomplished with minimum expenditure.
Some of that is positive; matching local installers to each day’s assignments to minimize travel time. But much of the potential for businesses is negative for workers. Any software allocating hour-by-hour staff that is prioritized around managing labor costs (as they generally are) can be perpetually figuring out the lowest pay that keeps sufficient work-seekers in play.
It could simultaneously be breaking down what patterns of hours, or allocation of shifts, give it an edge over any regulator trying to cap its cost-cutting power. We have spent years working to decode today’s opaque, incredibly sophisticated, blue-collar labor platforms. Takeaway: their capacity to cut hours and earnings is probably worse than most people can imagine. AI puts this on steroids.
----
Innovators will, of course, come up with AI tools designed to help workers fight all-powerful algorithms. Those apps should work, but they come up against platform operators’ need to maintain their one-sided market power. Remember UberCheats? Or Rodeo? Not many do. They’re in a graveyard of well-intentioned attempts to use tech to rebalance today’s labor markets that were quietly snuffed by the platforms.
Whac-a-mole regulatory attempts to stamp out abuse of platform-mediated workers will also continue. But AI adds urgency to another way of seeing the market for soft skills. Think of the problem from each individual’s point of view as siloing. I may work as an HVAC installer for an employer who is increasingly likely to use me as needed, rather than employ me full time. But I will have all sorts of other lived experience, and perhaps credentialing, from which I could also earn.
Maybe I can fit solar panels, paint houses, fit furniture and do landscaping. I might additionally be a short training course away from bricklaying, scaffolding or framing on building sites. If I could also be exposed to demand for all that potential, my HVAC employer would be forced to compete for every block of my hours they wished to book.
It’s often assumed that the blue-collar workforce operates in a market where each person has the information they need, and easy market access, to offer themselves across a range of possible sources of earnings then gravitate towards their best opportunities. It’s not true: if you are unavailable when Platform A needs you because you are doing a task sent by Platform B, A’s algorithms will downgrade you for unreliability, curtailing future work. And vice versa. You have to pick two or three of these task-based siloes and hope they treat you well.
The core issue, as our regular readers know, is that the thousands of platforms through which regular people now earn are not focused on running a supportive, progressive - or even fair - labor market. Each is striving for competitive advantage and to attract buyers of labor, knowing workers must go wherever the demand is.
----
Thus, AI is one more factor heightening the need for horizontal markets for hour- by- hour labor; platforms in which anyone can seamlessly offer their entire range of competencies, experiences and aspirations across a regional economy on their own terms. We are working to spread the word on just how much AI could empower people across the base of the economy when the bodies with sufficient heft to instigate all-sectors hourly labor markets, decide it is time to do so.