Definition of “superintelligence”
By a “superintelligence” we mean an intellect that is much smarter than the best human brains in practically every field, including scientific creativity, general wisdom and social skills. This definition leaves open how the superintelligence is implemented: it could be a digital computer, an ensemble of networked computers, cultured cortical tissue or what have you. It also leaves open whether the superintelligence is conscious and has subjective experiences.
Entities such as companies or the scientific community are not superintelligences according to this definition. Although they can perform a number of tasks of which no individual human is capable, they are not intellects and there are many fields in which they perform much worse than a human brain – for example, you can’t have real-time conversation with “the scientific community”.
This definition was written in 1998, but it seems to still capture a basic consensus of what we are looking for with our hopes for superintelligence. The definition that is given here excludes a superintelligence in one, closed example of meatspace. Intelligence no longer resides in the brain. Indeed, with each new accretion of technology in the world, a process of technological sedimentation covering all of our social reality in layer after layer of processes and techniques and tools to accomplish tasks that our mere corporeal beings cannot.
Vernadsky called this the noosphere, Kevin Kelly calls it the technium. Again, the fundamental point is that we have limits and we become aware of them (by our inability to do something), and so we design a way to step beyond our limits. This we call technology. So far, we have been quite successful in doing so. Computers are not intelligent by any means. They are still just an extension of our intelligence, a way for us to overcome our own limits.
I cannot see any alternative to presenting how things stand other than to say that the “entities” of which Bostrom writes above are the analogue, meatspace equivalent of computers that have served us well to this point. I also wonder if we fully appreciate the role of such social computing and calculation. If we did, would it change our view of what superintelligences might possibly be?
The mention of “real time” is also interesting here, because Bostrom has set up a division according to a timescale. This might have interesting implications for the definition of intelligence, however, because no digital computer will ever operate at the same speed as the human brain unless we fundamentally restrict it. If the human brain has a raw computational power between 1013 and 1016 operations per second (Bostrom gets 1014), and computers today are already at this stage, then surely with technological development in hardware and networking techniques, by the time (I prefer “if”) computers come to a level of sentience, then their “mental” operations will take place at speeds a number of orders of magnitude greater than our own. For them, talking to a human would be as slow as us “talking” to the scientific community (Bostrom says this doesn’t happen, but I consider the peer-review process, and academic fashion to be examples). Or it will be like talking to a child, then a dog, then a bird, then an insect, then a bacterium, until… Maybe we should think less in terms of sci-fi and more in terms of fantasy. Will we not seem like some creatures of old, speaking and thinking so slowly. What springs to mind for me here is not cyberpunk, but the Lord of the Rings: we will be like the tree ents. Treebeard, not Johnny Mnemonic. Slow, ponderous, but without the gravitas or power.