The Singularity Is Near!

What I had heard about the Singularity before reading Ray Kurzweil‘s book “The Singularity Is Near” was this: In the very near future computers will reach the processing power of the human brain, will become conscious and able to improve themselves, thereby getting exponentially smarter and quickly “taking over the world”.

I never doubted that computers would reach the human brain’s processing power, but I was always curious why the Singularity assumes that they’d automatically become conscious. Surely you need the right software for that!

Ray Kurzweil makes it very clear in his book what is needed for that step to take place, how to achieve it, and why he believes that we will achieve it very soon (in about 30 years). His main argument, for which he presents a great deal of quite convincing evidence, is that progress in science and technology is not linear, but exponential. That is especially true for the three key technologies enabling the Singularity, namely genetics, nanotechnology, and robotics.

In short, Kurzweil expects that we will come to a point where we will be able to scan a human brain, most probably via nanotechnology, and then “run” it on a computer, effectively creating an electronic copy of the brain’s proprietor. Even before that we will enhance our brains and bodies via genetics and nanotechnology, starting a transition from being purely biological creatures, merging with nanotechnology and robotics, to having purely digital brains and probably virtual bodies (physical bodies, whether biological or robotic, will be a hindrance once our speed of thought has accelerated by a few orders of magnitude).

What to think of all this? I didn’t doubt that we will at some point have the technology to “transcend biology”, as Kurzweil puts it, but I didn’t think that it might happen in my lifetime. Kurzweil’s vision depends on exponential progress in three technologies, and he presents very convincing arguments that not only are they currently progressing exponentially, but also that this exponential progress is sustainable at least up to the point where super-human machine-intelligence will be a reality, and at that point, further progress will be driven more and more by this superior intelligence. Am I completely convinced? No, but I’d say that it looks to be a real possibility. Unless we blow ourselves up first…

And if you’d like to get a fictional picture of what life after the Singularity might be like, I can recommend no better book than Greg Egan’s “Diaspora“.

5 thoughts on “The Singularity Is Near!

  1. I haven’t read that book, but I am aware of “singularity evangelism.” Please correct me, if you think am wrong, but my view is pretty simple.

    First, quite obviously, raw computational power does not suffice for producing intelligence.

    Second, I do not think that simulating the human brain on a functional level has any chance of success. If Kurzweil predicts that this simulation will be possible in 2013, we should have gone part of the road that leads there.

    Third, simulation on the molecular level may be the only option. But, then again, I am sceptical about scanning an existing brain and putting it into a computer. IMO it should die on the spot (because of deprivation of proper sensory inputs/outputs). And growing a brain from scratch inside the computer (based on simulations of the molecular level) is a task just as hard, presumably harder.

    Fourth, even if the required technological advances are possible, production costs are also rising dramatically. Gordon E. Moore predicted that in the 1970s. So, financial reasons (like increased production costs or some possible future decline in market demand) and not technical reasons might turn out to be the biggest obstacles to reaching the “singularity.”

    Still hoping you enjoyed reading the book…

  2. > First, quite obviously, raw computational power does not suffice
    > for producing intelligence.

    Nobody said it was.

    > Second, I do not think that simulating the human brain on a
    > functional level has any chance of success. If Kurzweil predicts
    > that this simulation will be possible in 2013, we should have
    > gone part of the road that leads there.

    Kurzweil predicts that we will reach the point where we will have the computational power for a functional simulation (for 1000$) of the brain around 2025. Simulation at the neuron level would have to wait until sometime in the 2030’s.

    He also gives several example of brain reverse-engineering that has already been done, for example Lloyd Watts’s work on the human auditory system.

    > Third, simulation on the molecular level may be the only option.

    Why would that be?

    > IMO it should die on the spot (because of deprivation of proper
    > sensory inputs/outputs).

    If you can simulate a complete brain it should be a minor matter to give it proper input and relay its output. A lots of successful work on this is being done today.

    > Fourth, even if the required technological advances are
    > possible, production costs are also rising dramatically. Gordon
    > E. Moore predicted that in the 1970s. So, financial reasons
    > (like increased production costs or some possible future decline
    > in market demand) and not technical reasons might turn out to be
    > the biggest obstacles to reaching the “singularity.”

    Kurzweil always uses “computational capacity for 1000$” as his source for predictions. But even if the market should decline, all it would do is delay the Singularity.

  3. Surely, simulations on several levels are conceivable.

    I doubt that the functional level will EVER succeed — no matter what finite computing power you have available — because of the complexity of the issue.

    I am not sure if a simulation on the level of neurons has much better chance of success either. Brains consist not only of neurons, but also of lots of other stuff. I cannot possibly believe that one can do without them: evolutionary processes usually do not “spend” too much energy on useless crap.

    Simulation on a molecular level might be an option, but it’s enormously compute intensive.

    In this particular context, I do not even want to use the term “singularity.” Even if the simulation of human brains inside computers will someday be feasible, we cannot expect these simulated brains to give any reasonable answer to the fundamental questions — at least any better answers than we can today. And it’s these questions these systems would need to answer for the “singularity” to occur.

  4. I’m skeptical of the singularity argument.

    If you’ve done any AI programming or computer science, it’s quite easy to differentiate between the kinds of intelligence a computer can possess vs. those of a human.

    Can a computer feel?

    If you define “feeling” such that the artificial feelings of a computer count, then you have to lump humans into the category of robots… i.e. we are just preprogrammed at birth and have no real free-will, but are all rather just deterministic, not unlike AI programs.

Comments are closed.