Desunt Cetera

A lot of bits have been spilled talking about the singularity, but I very often find myself with the impression that some of the contributors are talking at cross currents. So in this paper, I would like to lay out several conceptions of what the singularity is, starting with something that’s fairly uncontroversial, and progressing towards conceptions that cause some to go apoplectic. I do not wish to defend any one in particular, but rather layout three different conceptions of the event.

At its heart, the idea behind the singularity rests on a simple assumption: the rate of change of the rate of technological change is accelerating, and has been doing so as far back in time as we can discern.

The way to unpack this idea is to ask yourself, how far back in the past do you have to go before the everyday life of someone in a similar socio-economic situation is so different from your own that they would not be able to recognize it? For example, I use a personal computer for more hours a day than some would consider healthy. I only have to go back 20 or 30 years before I reach the point where a major part of my life is simply gone. So in a certain sense, to someone living in 1970, my day to day life is, if not incomprehensible, at the very least very strange.

Now, for that person in 1970, how far back must THEY go to get a similar result? I would argue that the boundary goes up to 30 - 50 years, i.e. their grandparents’ lifetime.

Lets look at it in the other direction. Imagine going back to Socrates, who lived around 400 BCE. If he were sitting around in the forum, trying to imagine what the future would be like, how far forward could he reasonably predict? I would argue he could say “for the next 1,000 years, people will live pretty much as I do,” and he would be right. Would you be willing to claim, today, that people 1,000 from now, will have a daily life substantially similar to your own?

So one way of cashing out the singularity is to say that the span of time over which we can make predictions about the daily life of ordinary individuals (with any hope of being near the mark) has gotten shorter over time. At some point, that span of time will be so absurdly short (5 years? a year? months?) that, for all practical purposes, we will not be able to make that sort of prediction at all.

This is the weakest way to formulate it, and probably the easiest to defend.

A stronger claim is that as a direct result of the accelerating rate of change of technology, certain types of technology become inevitable that will fundamentally alter human society. There are several different types of technology that may fall under this, but they generally have to do with granting human immortality in one form or another.

Proponents of this claim generally argue that one of the following scenarios will occur:

We will eliminate aging. At this time, we have only a basic grasp of what causes aging. It is hoped that eventually a “cure” for aging will be found, allowing people to live indefinitely. Note, however, that this does not prevent deaths from accident or injury.

We will develop cloning technology that lets us replace body parts at will. With this technology, rather than treating aging as a disease in and of itself, we would be able to replace organs with fresh new ones as they break down. As above, this would allow people to live indefinitely, but it also offers the hope of eliminating deaths from injuries.

We will develop artificial limbs and organs that are better than natural ones. Rather than replace a failing liver with a cloned one, we might replace it with a more reliable mechanical one that is not susceptible to disease. Some, like the Extropians, would want to apply this technology to healthy humans, arguing that there’s no reason to limit the use of such technology to make people “normal,” when they could also be used to make people super-human.

We will develop the ability to upload human consciousnesses to computers. Here, the idea is that we could abandon our human bodies entirely, and continue living in a computer simulation. I could write a whole essay on the subject, as there are numerous arguments about what exactly this means, and whether or not the resulting computer program is actually identical to the original person. I don’t have space to explore it further here, so I will just say that I myself believe that “I” am no more than a collection of experiences, thoughts and ideas. It shouldn’t make a difference whether that collection is instantiated in a brain or a computer or what have you. There are problems with this position, that I will address in a later post.

The strongest version of the last scenario, where all humans upload themselves into computers and abandon any physical instantiation, is sometimes derisively called the “Nerd Rapture” for its obvious parallels with the Christian Rapture.

An obvious problem with this formulation of the singularity is that it’s not a necessary fact that progress occurs. Just because it’s been speeding up over the past few hundred years does not give us reason to assume that this rate of change will continue to accelerate.

The strongest version of the singularity can be formulated in several different ways, as was done by Venor Vinge in his 1993 article “The Singularity Is Near”, which coined the name and introduced the idea. He cataloged three different ways the singularity could occur. They all revolve around the creation of super human intelligences, but differ in how it is done.

The first formulation is that we will develop intelligence amplifying technology that enhances human intelligence beyond what is capable of naturally. Perhaps we will develop technology that lets us connect computer hardware directly to human brains, drastically increasing our memory capacity, calculating ability, etc. If the increase in intelligence is great enough, problems that are hard for “normal” humans could become trivially easy to solve.

The second formulation is that perhaps we will develop technology that lets us create super-humanly intelligent humans “naturally.” In this scenario, we would use genetic engineering to create uber mench. Imagine a world where humans with intellects that surpass the most brilliant “normal” human to have ever lived are so common place as to be unremarkable.

The third formulation is that we may develop an artificial intelligence that is either smarter than humans, or as smart as a human but thinks exponentially faster.

If the AI is smarter than humans, it is reasonable to assume that it will be capable of developing another AI that is even smarter than it is, which will then be able to do the same. The justification for this is that we’ve already imagined that humans created an intelligence smarter than themselves, and one of the things human intelligence was capable was the creation of a new and greater intelligence. Ergo, the new AI can do the same, because it can by hypothesis do everything a human can do, as well as some things we cannot.

If the AI is only as smart as a human, but thinks exponentially faster, it may be able to create a smarter intelligence as in the previous case. Even if it cannot, the simple fact that it is able to think faster has profound implications for the development of technology. Imagine an inventor who, instead of taking months or years to perfect an idea, can do it in mere seconds.

In any of these three cases, Vinge claims that the resulting technology will so fundamentally alter the nature of technological progress that our ability to imagine the consequences is simply not up to the task. In effect, that date the first AI comes online is a technological singularity, after which we simply have no idea what the world will be like. Notice that this subsumes the previous formulations, as it is assumed that that these AIs will assist us in developing the technologies previously referenced.

Now, there are problems with each of these scenarios, but the one most often criticized is the third. Some argue that thinking machines, or at least conscious ones, are fundamentally impossible. See, for example, What Computers Can’t Do by Hubert Dreyfus, or the literature surrounding the Chinese Room argument by Searle for a small taste of the raging arguments for and against the possibility.

Vinge makes some strong claims in his essay, including that the singularity is inevitable. As noted above, this has earned him criticism by those who point out that technological progress is not a necessary fact, and we could very easily destroy ourselves long before we reach the singularity. I’m not going to defend or criticize Vinge in this essay, as it was not my goal. I’ll limit myself to saying that his paper is well worth reading, as is its more modern equivalent, the book The Singularity Is Near by Raymond Kurzweil.

It is my hope that those who read this will be more careful when they use the term “singularity”, and specify what exactly they are talking about when they do so. To wit, do they mean 1) our ability to predict the future gets more and more circumscribed as technology advances, or 2) that that at some point, probably sooner rather than later, some technology will be developed that will so alter humanity that we will become post human or 3) that we shall soon develop sophisticated AI that will cause an intelligence explosion unlike anything we have ever experienced before. These are different claims, with different causes, effects and criticisms.

Philosophy