An organization could be viewed as a type of mind with extremely redundant modular structure. Human minds contain a large number of interconnected specialized subsystems, in an organization humans would be the subsystems. Comparing the two seems illuminating.
Individual subsystems of organizations are much more powerful and independent, making them very effective at scaling and multitasking. This is of limited value, though: it mostly just means organizations can complete parallelizable tasks faster.
Intersystem communication is horrendously inefficient in organizations: bandwidth is limited to speech/typing and latency can be hours. There are tradeoffs here: military and emergency response organizations cut the latency down to seconds, but that limits the types of tasks the subsystems can effectively perform. Humans suck at multitasking and handling interruptions. Communication patters and quality are more malleable, though. Organizations like Apple and Google have had some success in creating environments that leverage human social tendencies to improve on-task communication.
Specialization seems like a big one. Most humans are to some degree interchangeable: what one can do, most others can do less effectively, or at least learn given time. There are ways to improve individual specialization, but barring radical cultural or technological change, we’re pretty much stuck on that front.
Mostly organizations seem limited by the competence of their individual members. They do more, not better. Specialization and communication seem to be the limiting factors and I’m not sure if they can make enough of a difference even in theory to qualify as a superintelligence, except in the sense a sped-up human would.
One of the advantages of bureaucracy is creating value from otherwise low-value inputs. The collection of people working in the nearest McDonalds probably isn’t capable of figuring out from scratch how to run a restaurant. But following the bureaucratic blueprint issued from headquarters allows those same folks to produce a hamburger on demand, and getting paid for it.
That’s a major value of bureaucratic structure—lowering the variance and raising the downside (i.e. a fast food burger isn’t great, but it meets some minimum quality and won’t poison you).
I think that you are underestimating the efficiency of intersystem communication in a world where a lot of organizational communication is handled through information technology.
Take a modern company with a broad reach. The convenience store, CVS, say. Yes, there is a big organizational hierarchy staffed by people. But there is also a massive data collecting and business intelligence aspect. Every time they try to get you to swipe your CVS card when you buy toothpaste, they are collecting information which they then mine for patterns on how they stock shelves and price things.
That’s just business. It’s also a sophisticated execution of intelligence that is far beyond the capacity of an individual person.
I don’t understand your point about specialization. Can you elaborate?
Also, I don’t understand what the difference between a ‘superintelligence’ and a ‘sped-up human’ would be that would be pertinent to the argument.
I think that you are underestimating the efficiency of intersystem communication in a world where a lot of organizational communication is handled through information technology.
Speech and reading seem to be at most 60 bits per second. A single neuron is faster than that.
Compare to the human brain. The optic nerve transmits 10 million bits per second and I’d expect interconnections between brain areas to generally fall within a few orders of magnitude.
I’d call five orders of magnitude a serious bottleneck and don’t really see how it could be significantly improved without cutting humans out of the loop. That’s what your data mining example does, but it’s only as good as the algorithms behind it. And when those approach human level we get AI.
I don’t understand your point about specialization. Can you elaborate?
Individual humans have ridiculous amounts of overlap in skills and abilities. Basic levels of housekeeping, social skills etc. are pretty much assumed. A lot of that is necessary given our social instincts and organizational structures: a savant may outperform anyone in a specific field, but good luck integrating them in an organization.
I’m not sure how much specialization can be improved with baseline humans, but relaxing the constraint that everyone should be able to function independently in the wider society might help. Also, focused training from a young age could be useful in creating genius-level specialists, but that takes time.
Also, I don’t understand what the difference between a ‘superintelligence’ and a ‘sped-up human’ would be that would be pertinent to the argument.
Given a large enough speedup and indefinite lifespan, pretty much none. The analogy may have been poorly chosen.
Wait...one sec. Isn’t all that redundancy in human society a good thing, from the perspective of saving it from existential risk?
If I were an AI, wouldn’t one of the first things I do be to create a lot of redundant subsystems loosely coordinating in some way, so that if have of me is destroyed, the rest lives on?
It looks to me like there’s a continuum within organizations as to whether they do most of their information processing using hardware or wetware.
I acknowledge that improvements in machine intelligence may shift the burden of things to machines.
But I don’t think that changes the fact that many organizations already are superintelligences, and are in the process of cognitively enhancing themselves.
I guess I’d argue that organizations, in pursuit of cognitive enhancement, would coordinate their human and machine subsystems as efficiently as possible. There are certainly cases where specialists are taken care of by their organizations (ever visited a Google office, for example?). While there may be overlap in skills, there’s also lots of heterogeneity in society that reflects, at least in part, economic constraints.
Human minds contain a large number of interconnected specialized subsystems, in an organization humans would be the subsystems.
In a company large enough, the humans would be like the cells, and the departments would be the subsystems. The functional difference between e.g. the accounting department and the private security department can be big, even if both are composed of biologically almost the same homo sapiens individuals.
When comparing the speed of organizations with speed of humans, on different scales the speed comparison can be different. As an analogy, a bacterium can reproduce faster than a human, but a human will write a book faster. Similarly, humans can do many things faster than organizations, but some other things are just out of reach for an individual human without an organization of some kind.
I would say that today, humans are relatively advanced in the human-space, shaped by biological evolution and culture for a long long time. Compared with that, organizations seem rather primitive and fragile in the organization-space. Yet even today the organizations can do things that individual humans can’t. It is like looking at the first multi-cellular organisms and deciding that although they have some small advantages over the single-cellular ones, they are not impressive enough.
An organization could be viewed as a type of mind with extremely redundant modular structure. Human minds contain a large number of interconnected specialized subsystems, in an organization humans would be the subsystems. Comparing the two seems illuminating.
Individual subsystems of organizations are much more powerful and independent, making them very effective at scaling and multitasking. This is of limited value, though: it mostly just means organizations can complete parallelizable tasks faster.
Intersystem communication is horrendously inefficient in organizations: bandwidth is limited to speech/typing and latency can be hours. There are tradeoffs here: military and emergency response organizations cut the latency down to seconds, but that limits the types of tasks the subsystems can effectively perform. Humans suck at multitasking and handling interruptions. Communication patters and quality are more malleable, though. Organizations like Apple and Google have had some success in creating environments that leverage human social tendencies to improve on-task communication.
Specialization seems like a big one. Most humans are to some degree interchangeable: what one can do, most others can do less effectively, or at least learn given time. There are ways to improve individual specialization, but barring radical cultural or technological change, we’re pretty much stuck on that front.
Mostly organizations seem limited by the competence of their individual members. They do more, not better. Specialization and communication seem to be the limiting factors and I’m not sure if they can make enough of a difference even in theory to qualify as a superintelligence, except in the sense a sped-up human would.
Thoughts?
One of the advantages of bureaucracy is creating value from otherwise low-value inputs. The collection of people working in the nearest McDonalds probably isn’t capable of figuring out from scratch how to run a restaurant. But following the bureaucratic blueprint issued from headquarters allows those same folks to produce a hamburger on demand, and getting paid for it.
That’s a major value of bureaucratic structure—lowering the variance and raising the downside (i.e. a fast food burger isn’t great, but it meets some minimum quality and won’t poison you).
I think that you are underestimating the efficiency of intersystem communication in a world where a lot of organizational communication is handled through information technology.
Take a modern company with a broad reach. The convenience store, CVS, say. Yes, there is a big organizational hierarchy staffed by people. But there is also a massive data collecting and business intelligence aspect. Every time they try to get you to swipe your CVS card when you buy toothpaste, they are collecting information which they then mine for patterns on how they stock shelves and price things.
That’s just business. It’s also a sophisticated execution of intelligence that is far beyond the capacity of an individual person.
I don’t understand your point about specialization. Can you elaborate?
Also, I don’t understand what the difference between a ‘superintelligence’ and a ‘sped-up human’ would be that would be pertinent to the argument.
Speech and reading seem to be at most 60 bits per second. A single neuron is faster than that.
Compare to the human brain. The optic nerve transmits 10 million bits per second and I’d expect interconnections between brain areas to generally fall within a few orders of magnitude.
I’d call five orders of magnitude a serious bottleneck and don’t really see how it could be significantly improved without cutting humans out of the loop. That’s what your data mining example does, but it’s only as good as the algorithms behind it. And when those approach human level we get AI.
Individual humans have ridiculous amounts of overlap in skills and abilities. Basic levels of housekeeping, social skills etc. are pretty much assumed. A lot of that is necessary given our social instincts and organizational structures: a savant may outperform anyone in a specific field, but good luck integrating them in an organization.
I’m not sure how much specialization can be improved with baseline humans, but relaxing the constraint that everyone should be able to function independently in the wider society might help. Also, focused training from a young age could be useful in creating genius-level specialists, but that takes time.
Given a large enough speedup and indefinite lifespan, pretty much none. The analogy may have been poorly chosen.
Wait...one sec. Isn’t all that redundancy in human society a good thing, from the perspective of saving it from existential risk?
If I were an AI, wouldn’t one of the first things I do be to create a lot of redundant subsystems loosely coordinating in some way, so that if have of me is destroyed, the rest lives on?
It looks to me like there’s a continuum within organizations as to whether they do most of their information processing using hardware or wetware.
I acknowledge that improvements in machine intelligence may shift the burden of things to machines.
But I don’t think that changes the fact that many organizations already are superintelligences, and are in the process of cognitively enhancing themselves.
I guess I’d argue that organizations, in pursuit of cognitive enhancement, would coordinate their human and machine subsystems as efficiently as possible. There are certainly cases where specialists are taken care of by their organizations (ever visited a Google office, for example?). While there may be overlap in skills, there’s also lots of heterogeneity in society that reflects, at least in part, economic constraints.
In a company large enough, the humans would be like the cells, and the departments would be the subsystems. The functional difference between e.g. the accounting department and the private security department can be big, even if both are composed of biologically almost the same homo sapiens individuals.
When comparing the speed of organizations with speed of humans, on different scales the speed comparison can be different. As an analogy, a bacterium can reproduce faster than a human, but a human will write a book faster. Similarly, humans can do many things faster than organizations, but some other things are just out of reach for an individual human without an organization of some kind.
I would say that today, humans are relatively advanced in the human-space, shaped by biological evolution and culture for a long long time. Compared with that, organizations seem rather primitive and fragile in the organization-space. Yet even today the organizations can do things that individual humans can’t. It is like looking at the first multi-cellular organisms and deciding that although they have some small advantages over the single-cellular ones, they are not impressive enough.