Unemployment — Is This Time Different?
We’re hearing much Doomsday talk around technological unemployment, prompting increased support for some form of Universal Basic Income.
On the other hand, we have others claiming that there’s nothing to worry about: Luddite fallacy! We’ve always had new jobs come up as others are automated! It’s Econ 101! Read up on comparative advantage!
What there does seem to be agreement on is:
- Automation will continue to replace existing jobs
- There will be ‘dislocations’ — a polite way of saying ‘people losing their jobs, and needing to change profession’
- Artificial Intelligence and robotics are driving this trend
- It will probably do so at an accelerating pace
The following analysis makes the explicit assumption (not defended here) that full human-level Artificial General Intelligence (AGI) will become available over a time period of less than 20 years. Current mainstream AI technology is not yet on a trajectory towards that goal.
Human-level AGI implies that machines will be able to perform almost all tasks as well, or better than humans. To be clear, we’re talking about a world where robots are better at designing, building, and servicing robots than humans are.
Let me qualify this by acknowledging that many of us will at times prefer to be served by a (more expensive, and probably less competent) human rather than by a robot — be it for snob value, nostalgic impulse, or some perceived additional quality.
To put this into perspective though, we should remember that AGI’s will inherently have several significant advantages over humans:
- Can work accurately 24/7 without getting bored or hungry
- Can easily share information, skills, and insights between them
- Once trained, can easily be cloned/ copied
- Don’t suffer large egos or other ‘reptile brain’ drives
- Are not distracted by the opposite (or same!) sex
- Can have photographic memory (as needed)
- Can easily be sped up significantly for many tasks
- Have immediate ‘mental’ access to all online information
- Have much better reasoning skills, and fewer harmful biases
- Can recognize and respond to emotions better than humans
So the situation that unfolds is that over a relatively short period (less than 20 years) more and more jobs will be replaced by much less expensive and more capable robots.
Furthermore, and importantly, the robot’s cognitive and physical skill levels will keep increasing, reducing the pool of humans that are even capable of performing the remaining human tasks at any one time.
Most of the ‘new jobs’ (goods and services) that are created will optimized for and integrated with the AGI infrastructure, and not suitable for humans.
So why can’t people just offer their labor for tasks appropriate to their skills at a rate lower than the cost of the robots?
There are several reasons:
- The automated economy will increasingly be designed for robots — relying on their inherent strengths, and humans will simply not fit in. Can humans replace a computer in a fully autonomous car, or factory robot? Or travel websites algorithms? Or a robot surgeon?
- More generally, for jobs that can be done cheaply by a robot (most of them), humans will actually have a disvalue because of their inherent inconsistency, reliability, personal issues, etc. — similar to the way that many people today are not worth hiring at any cost for particular jobs. In the future, who would want to hire a human programmer, and deal with their limitations? At any price?
- Training is another huge issue. In the abundance economy brought about by automation, most people will need to work much less. This creates the problem of humans not having current skills (after a long time off), or the additional cost of having to train and manage a large number of part-time workers, who may only work a few hours a week.
- Robots will make hiring decisions :(
So if most humans will essentially be unemployable, will we all be destitute?
I was recently introduced to the concept of Shrödinger’s Robot (by Fare) — essentially the paradoxical fear that robots will simultaneously be so cheap as to force us out of the labor market, and yet us being too poor to afford the ‘too-cheap-to-meter’ services they provide. (He jokes that these robots will be built by Shrödinger’s immigrant.)
The bottom line seems to be that advanced AGI will indeed make most human jobs obsolete — freeing us up to discover new and better ways of flourishing.
In time, we will have to learn how to cope with our new-found freedom, and we’ll have to figure out how to benefit from our robot-created abundance equitably. Fortunately, we’ll have AGI to help us figure that out.