Recently, I read an article in Time Magazine titled 2045: The Year Man Becomes Immortal In it, they discuss the rapid advance in artificial intelligence, and have popularized the phrase "The singularity" as the moment when computers become capable themselves of designing more intelligent computers ad infinitum.
Time Magazine did not coin the term "Singularity" as it pertains to AI.
So-called "post-humanism" has driven a large amount of fiction as authors speculate as to the social impact of the introduction of a Artificial General Intelligence (the term "general" is important; artificially intelligent programs already exist, but are very specific in their application, not adaptable to multiple or unforeseen situations). We're all familiar with the Terminator-style apocalyptic visions of the effects of AGI, for example.
As to whether a Singularity event will happen at all...it seems inevitable that AGI will eventually be developed. After all, we know that a general intelligence is possible - every human brain is an example of an adaptable intellect. It seems logically foolish to claim that while a natural general intelligence is possible, it is impossible for humans to artificially create one.
It's far, far easier to modify a computer program or computer hardware than it is to modify the living brain of a human being.
It's far, far easier to keep track of what specifically is going on inside of a computer program, what task is currently being processed, and how specifically it works, than it is to do the same with a living human brain.
Computer processing happens
far faster than human thought, and further is not limited to a specific amount of space. A human brain needs to fit within the confines of a human skull, and has to work with the finite energy and nutritional resources of a human body. Electronic computers are far more easily up-scaled, adding additional processors or memory or storage, etc.
Can you imagine, then, an artificial intelligence that is capable of altering itself? Of analyzing its own thought processes, assessing its own performance pursuant to its goals, and modifying its core programming to be more efficient? Human beings are flawed intelligences, with cognitive faults like confirmation bias; imagine being able to actually, literally,
reprogram those faults out of your thought process. An AGI would likely be able to do that, and depending on the hardware allocated and processing requirements for running the program, might well be able to do it faster than we can track.
Some of the implications of AGI are difficult to predict; others are easy.
Some potential pathways to AGI involve trying to duplicate a human brain in a computer environment; one proposed technique involves attempting to simulate every neuron in a human brain within a computer. This might make a more human-like AGI, but has significant downsides - indirect simulation processing is relatively inefficient, and simulating human neurons is rather like copying Chinese when you can't read the language - you still have very little idea what's going on inside (conversely, one of the benefits is that such a simulation, since we could pause the simulation and trace the activity of every neuron, might actually help us gain a better understanding of our own brains). The religious reaction to such a human-like AGI seems easy enough to predict - there would be debates over whether such constructs have "souls," whether they are an example of "playing God," and so on.
Imagine that neuron-simulation AGI paves the way for human brain uploads - being able to transfer a human consciousness from a biological body to a computer. Would "uploaded" humans still retain human rights? If, once uploaded, such intelligences can modify themselves, would they tend to remain particularly "human" in their thought processes? Would they simply simulate for themselves a Utopian dreamworld like the Matrix and leave the rest of us behind? If survival simply means an available power source and an industrial base capable of replacing defective components, and if
backups are possible, would this then not be a route to human immortality?
Other potential pathways involve actually creating an AGI from the ground up, not trying to copy the human brain at all. Some of these methods carry the benefit of being able to better understand what's happening in the AGI and how it works from day one, but may carry the downside of being wholly alien to the way we think. There's no requirement that an AGI be sentient, or contain self-preservation as a goal, and so on. Interacting with a machine that is quite literally
smarter than you but is not self-aware in the way that you or I am, or which has such drastically different goals than humans generally follow (self-preservation, procreation, entertainment, etc) would be an
interesting experience, to say the least. An "alien" AGI would likely face a steeper battle for any sort of rights; an AGI that isn't sentient would likely not even trigger much ethical or religious debate at all.
If an AGI is simply a complex computer program, it would be trivially simple to
copy an AGI. Creating just one means that it should be possible to create as many as you like provided you can obtain the hardware to run them.
AGIs, not being bound to the requirements of flesh-and-blood bodies, could open up new horizons for space travel. A 1000-year long trip sounds less troublesome for an AGI whose lifespan would be measured in terms of available power and the occasional replacement part. Food, water and air would be unnecessary, as would exercise, space to move around, some radiation shielding, etc. An AGI could directly interface with all manner of sensors, not be limited to the senses of a human being; it could directly see x-rays rather than using false-color intermediary photographs, for example. A single AGI on a ship equipped with telepresence utility bodies (repair bots, rovers, probes, etc) could hypothetically carry out an entire interplanetary mission on its own. All it needs is enough fuel for transportation and power.
The Singularity, at its heart, represents the question, "what happens when we build something that's smarter than we are?" Will we be left behind as our intellectual children outpace us? Will we "upgrade" ourselves with cybernetic implants or upload our brains to keep up? Will the machines be friendly towards us, or will they try to kill us off as potential threats?
It's easy to predict things like easier space exploration given AGI. It's easy to predict some cultural challenges, like rights for AGIs, their place in society, religious reactions, etc.
It's difficult to predict exactly what happens to
us. Many post-humanists base their predictions on the assumption that AGI would be able to solve many of our basic problems, like the energy crisis, through an exponential increase in technological advancement. While in general that assumption isn't too unlikely, the
specifics are where they start to fail. The development of an AGI tomorrow doesn't necessarily put us on a 3-year plan to working fusion power reactors. If human "uploads" become possible, that says nothing about the
cost of such a procedure or the hardware to run it.
I don't think we can accurately make predictions about the social consequences, simply because there are too many massively important variables. What is the AGI like? What are its core goals, and how do we fit into them? Can it
change its goals? Would it? Can we reasonably pull the plug? What's the economic cost of running one? Will we be limited to only a few in the world because they need building-sized data centers and lots of power, or will they be as common as personal computers? Can human beings relate to them in our interactions, or are they so much different in their thought processes that they feel alien to us? Are they sentient?
That's why opinions of the Singularity are so varied. It's very likely that an AGI will be developed; it's very likely that an AGI could be our intellectual superior in every way, thinking faster, recalling data more accurately, being able to modify itself to overcome flaws without lengthy 12-step programs, and able to integrate new information quickly and easily. Everything beyond that, from HAL-9000 to the Culture to Terminator to the Matrix to a post-humanist Utopia is nearly pure speculation.
™ Version 4.2
(1)
