Currently, we have software that can write software. What about physical machines that can produce physical machines? Indeed, what about machines that can produce other machines without human direction?
First of all, machines-building machines (MBM) still require resources: energy, transportation, time, and other inputs. A well-programmed machine that self-replicates quickly can grow in number exponentially. But where would the machines get the resources that enable self-replication? They’d have to purchase them (or conquer the world sci-fi style). Where would a machine get the resources to make purchases of necessary inputs? The same place that everyone else gets them.
In order for machines to gather resources that are pertinent to the production of other machines, external resources are necessary. Indeed, it’s what humans do. We create one type of resource in a specialized fashion, then we convert it into other resources for our consumption (broadly construed). MBM would have to do the same. If self-replicating machines only self-replicated, then there are 2 options: (1) Sell the replications & (2) Run out of resources and cease self-replication. But, if MBM can produce machines with a variety of functions, then part of the inputs that the MBM use will go toward self-replication and another part will go toward producing machines that create final goods and services. Without this second type of production, MBM can’t gather the requisite resources for replication. The result is a lower, though still exponential, rate of MBM population growth.
Indeed, if machines can be built at a vanishingly small marginal cost, and so can the production that is necessary to fuel the replications, then there are several implications.
- Owners of MBM will have a competitive advantage in production.
- Owners of MBM will enjoy large initial profits.
- The price of output & MBM will be driven down to near zero.
- The price of labor relative to its marginal product becomes less attractive to producers.
Start Saving Now: The profits earned by MBM will go to their owners – the owners of capital. You too can prepare for the coming world of super abundance by being one of those MBM owners. You’ll need resources when the time comes. Then again, maybe saving is superfluous when MBMs are also super-abundant and cheap.
Invest in Index Funds: The odds are very good that you will not be able to predict which company will create the first MBM. Given that the value of that company will very, very quickly go very, very high, even owning a small sliver changes the future path of your income schedule. Owning some of that MBM equity will insulate your from the negative shock that your wages will experience when labor is becomes a relatively less valuable input.
Treat your Friends and Family Well: Humans are social creatures. Happiness and flourishing require meaningful relationships. When MBM remove the opportunity for work-friends and reduce our physical dependance on co-producing domestic partners, our mental health will be contingent on whether others find us ‘good’ to be around. We need to provide meaningful relationships to others in order for them to stick around and provide the same to us. Uncomfortably, a world of superabundance might cause a wave of separations and divorce as unpleasant relationships, which only exist because of their economic benefits, lose their attractiveness. Why stay with a rich jerk when you are still unimaginably rich without them?
IOW: If you’re worried about ‘machines taking our jobs’ or excited for the machine-age revolution, then invest like there is a tomorrow and like you want to be a part of it.
*I’m leaving aside the prospect that MBM would trade with MBM. If that happens, then machines would not trade so beneficially with humans. We incur relatively high production and transactions costs. They’ll largely ignore us – hopefully without killing us while they do their thing.
1. The discussion on MBM reminds me of some eerie parallels with the economics of the slave economy in the US, particularly after the import ban. “Self-replicating capital” created an absurdly wealthy leisure class in the South, but didn’t the dominance of slave-dependent agriculture stunt growth in other sectors? By the same argument, wouldn’t the incredible profits of certain MBM-friendly sectors skew the economy for things that can’t be automated?
2. On the other hand, it seems that the bottleneck on production these days isn’t physical capital, it’s resources. If I wanted to double car production, I need twice as much steel, rubber, etc. On the other hand, it seems easier to just make another car factory. Creating a car factory *that also makes factories* seems like an unnecessary and inefficient step.
3. While your thoughts on hardware MBM might come true eventually, I note that software writing software seems much more likely to happen, much sooner. Software can iterate very quickly, and both inputs (cloud computation and storage?) and outputs (lines of code? biochemical pathways?? pulp novels???) don’t need any physical storage or transportation.
1) I agree. But, the MC of output production would be driven even lower than with slavery depending on the replication rate. I also agree that less replicable services would be produced by people instead. It depends on what we assume about the machine abilities.
2) I disagree strongly. We are using even less raw material now than we were 10-30 years ago. This is due to efficiencies, but also due to competing digital goods with a very low MC. Further, if people are paying for human digital producers, the price of labor generally will be higher – including raw material extractors (which has a higher MC). Finally, if the machines can extract raw materials at a very low MC, then there ceases to be a bottleneck. You say “easier to make another car factory” than to build a factory-building factory. This comment leads me to think that your missing the part about MC plummeting. *We can* produce factories slowly and expensively. MBM can building them quickly and cheaply at MC.
3) I 100% agree and I think that most other people do to. Which is why I didn’t write about it. Except, material services (experiences) demand will also increase. Absent convincing VR, the big question is whether AI & MBM can produce the social/experiential services that we’re speculating that they can’t.
2. I may have a conceptual disconnect here, because I’m thinking if the MC of building “factory that produces factories and cars” goes down, then wouldn’t the MC of building a “factory that only produces cars” and a “factory that only produces factories” also go down? Then it would always be more efficient to specialize in one thing.
Otherwise, by your analogy, I could extend it to “factory that produces factories and cars and also mines and refines ore and…” etc.
3. It’s funny, my parents’ generation are very much about collecting “stuff”, and my generation seems to be much more about “authentic experiences”, but my kids and their cohort are perfectly comfortable with virtual spaces. A YouTube video about an exotic location is almost as good as going there; a famous artist live-streaming a Fortnite server is almost as good as going to a “real” concert. I think they would be very accepting of AI-enabled experiences.
(And I’m sure my grandkids will completely rebel and accept only the most human-made, handcrafted experiences)