(These are excerpts from my book "Intelligence is not Artificial")
The Timeframe of Artificial General Intelligence
If by "artificial intelligence" we simply mean a machine that can do something (not everything) that we can do (like recognizing cats or playing chess), but not "everything" that we can do (both the mouse and the chess player do a lot of other things), then all machines and certainly all appliances qualify. Some of them (radio, telephone, television) are even forms of superhuman intelligence because they can do things that human brains cannot do.
Definitions do matter: there is no single answer to the questions "when will machines become intelligent" and "when will superhuman intelligence appear". It depends on what we mean by those words. My answer can be "it's already here" or "never".
As it stands, predictions about the future of (really) intelligent machines (of AGI) are predictions about a technology that doesn't exist. You can ask a rocket scientist for a prediction for when a human being will travel to Pluto: that technology exists and one can speculate what it will take to use that technology for that specific mission. On the contrary, my sense is that, using current technology, there is no way that we can create a machine that is even remotely capable of our routine cognitive tasks. The technology that is required does not yet exist. The machine that is supposed to become more intelligent than us and not only steal your job but even rule the world (and either kill us all or make us immortal) is pure imagination, just like angels and ghosts.
It is difficult to predict the future because we tend to predict one future instead of predicting all possible futures. Nobody (as far as i know) predicted that the idea of expert systems would become irrelevant in most fields because millions of volunteers would post knowledge for free on something called World-wide Web accessible by anybody equipped with a small computer-telephone. That was one possible future but there were so many possible futures that nobody predicted this one. By the same token, it is hard to predict what will make sense in ten years, let alone in 50 years.
What if 3D printing and some other technology makes it possible for ordinary people to create cheap gadgets that solve all sorts of problems. Why would we still need robots? What if synthetic biology starts creating alternative forms of life capable of all sorts of amazing functions. Why would we still need machines? There is one obvious future, the one based on what is around today, in which machines would continue to multiply and improve. There are many other futures in which computers and robots would become irrelevant because of something that does not exist today.
Anders Sandberg and Nick Bostrom, authors of "Whole Brain Emulation" (2008), conducted a "Machine Intelligence Survey" (2011) that starts with a definition of what an artificial intelligence should be: a system "that can substitute for humans in virtually all cognitive tasks, including those requiring scientific creativity, common sense, and social skills." My estimate for the advent of such a being is roughly 200,000 years: the timescale of natural evolution to produce a new species that will be at least as intelligent as us. If Artificial Intelligence has to be achieved by incremental engineering steps starting from the machines that we have today, my estimate about when a machine will be able to carry out a conversation like this one with you is: "Never". I am simply projecting the progress that i have witnessed in Artificial Intelligence (very little and very slow) and therefore i obtain an infinite time required for humans to invent such a machine.
But then, again, we'd probably have a lengthy discussion about what "all cognitive tasks" really means. For example, leaving out consciousness from the category of cognitive tasks is like leaving out Beethoven from the category of musicians simply because we can't explain his talent.
As i wrote, machines are making us somewhat dumber (or, better the environments we design for automation make us dumber), and there is an increasing number of fields (from arithmetic to navigation) in which machines are now "smarter" than humans not only because machines got smarter but also because humans have lost skills that they used to have. If i project this trend to the future, there is a serious chance that humans will get so much dumber that the bar for artificial general intelligence will be lower and therefore artificial intelligence more feasible than it is today; and "superhuman" intelligence may then happen, but it should really be called "subhuman" intelligence.
Back to the Table of Contents
Purchase "Intelligence is not Artificial"