http://theprogressive.wordpress.com/2006/09/19/kurzweil-is-wrong/

Kurzweil is wrong.

Hofstadter said it. R.U. Sirius certainly seems to feel it, while he has not come out and said it, to my knowledge. Kurzweil’s Law of Accelerating returns does not mean what Kurzweil argues it does. Here’s how it really works: Change is stochastic. Progress, as an epiphenomenon of change, is unpredictable. Kurzweil sets a hard date of 2045 on the Singularity. To put it bluntly: bullshit. Kurzweil, I don’t want to strawman you here… I know it’s not a hard date, more of an upper limit, but I really don’t think you’re properly understanding the process. Kurzweil’s limits are set upon the rates of telecommunication, which govern the rate of change, and the limits of computation, which to a certain extent govern the rate of human innovation, because computers can solve through stocastic or brute-force methods problems which are virtually random in terms of the time it takes for spontaneous human innovation. However, spontaneous human innovation presents a completely unpredictable factor in futurism, and progress, as an epiphenomenon of change, is inherently unpredictable, although more quickly converges upon the asymptote as the rate of change increases. Kurzweil’s Law does not represent an absolute metric of progress, only an absolute metric of the rate of change which only assists in the eventual convergence of progressive systems, but one which outside forces can throw askew at any time. Futurist predictions rely almost completely on predictions that after a certain amount of time innovation will inevitably emerge, but such an approach is a mistake. There are so many random, unpredictable components to the system that it simply becomes impossible to give anything a specific deadline. The abstractions that humans build upon are going to become increasingly complex, and the ability to predict the timescales of completion of such tasks, particularly the abstraction of the entire system of the universe into a human-created universal-universe abstractor system, something known as Artifical General Intelligence (AGI), become that much more unpredictable. Tinges of Revelation begin to peak their head out. Much like the rapture, no one will know the hour of AGI’s birth. The rate of change, accelerated by the increasing speed of the telecommunications network which drives the motivating factors for change at an ever-increasing pace, means that as time goes by the rate at which the invention of AGI could transform society only quickens. Its creation represents a trigger for sweeping change whose potential grows more profound with each passing day. The cause of this trigger remains spontaneous human innovation, but there are certain precursors we can look for which give definite assurances that AGI is near. The first would be a revolution in neurophysiology, specifically the creation of a comprehensive and falsifiable model of the operation of the human brain. To my knowledge there is but one individual working towards this goal, Jeff Hawkins, creator of the Palm Pilot, who has since founded the Redwood Neuroscience Institute with the goal of developing a comprehensive theory of the operating of the brain, and the company Numenta who seeks to implement in software his memory-prediction framework, a conceptual model of what many see as the three most important centers of the brain surrounding consciousness: the neocortex which represents the perceiver process, the thalamus which represents conscious state perceived by the neocortex, continuously altered by feedback loops between the two of them, and the hippocampus, the brain’s archivist which saves the state of these feedback loops to permanent memory. Hawkins hypothesizes that these three parts of the brain comprise the fundamental structures which represent consciousness in his book On Intelligence, which lays out the memory-prediction framework his company is attempting to develop. This potential path is also governed by the results of the Blue Brain Project, presently using the world’s most powerful supercomputer architecture to simulate the behavior of the fundamental unity of the rodent neocortex. The results of that study will give dramatic insights to those working on a comprehensive, scientific theory of the brain’s operation, including Jeff Hawkins. It’s potential that with enough empirical scientific research into and modeling of the functional operation of the human brain, the closer we will come to having the knowledge to implement artificial general intelligence. The second path involves increasing our intelligence of cellular and molecular biology to the point that we are able to create a computer simulation of the human developmental process from fertilized egg to human baby. In such a simulation the only limitations on the types of data we can collect are the error from imprecision of the model, the timesteps that the simulation is run over, and the storage limitations for the types of data that can be collected. The first point will likely not matter for any virtual baby which we can carry to term through a cellular model: due to the complex intracacies of the human developmental process, any simulation error apt to have an effect after the end of the virtual gestation period will likely have a negligable effect on the subsequent baby, otherwise the developmental process will fail and a human baby will never be constructed in simulation. Given the massive amounts of data that would be available from such a simulation, it’s likely the precise structure of human consciousness could be derived from the data using only analysis of the resulting simulation data. Thus the other harbinger of AGI you need to be on the look out for are computer models of multicellular organisms, which will eventually give way to full scale simulations of human beings. This potential path is governed by the results of the Blue Gene Project, presently using the world’s most powerful supercomputer to simulate the inner workings of the most complex and confounding aspect of intracellular behavior: protein folding. We are coming to understand it only through an atom-by-atom simulation of forces, particularly Gibbs Free Energy, which beomes one of the dominating factors in the folding pathways and kinetics of proteins. While I said before that human innovation is completely unpredictable, I believe all other paths to AGI will ultimately lead to failure. While I admit that there is the potential that someone taking an unscientific, completely conceptual approach to understanding the mind may leap frog everyone else by spontaneously coming up with the “right answer”, I believe the two horsemen of the technological infocalypse to bet on are a comprehensive theory of the brain and biomodeling. I also believe these two fields represent the two most beneficial scientific studies mankind can undertake, because I believe the fruits of their labor will eventually lead to AGI more than any other type of research. I’m sad to say that I do not believe that the MIT AI lab, the Singularity Institute, or any other group advocating direct research into AGI at the present time, or even advocating a non-biological approach towards successive steps towards AGI, will accomplish anything before the biological movement does, because while the former movement relies on the completely sporadic and unpredictable process of human innovation, the latter makes continous progress through scientific research into the problem. While the former will eventually usurp the results of the latter to create AGI, until that time comes we are floundering. There is much more we need to understand about nature before AGI becomes a possibility, and when the prerequisite natural knowlege is in place, we will make the leap to a post-AGI, post-Singularity society where were are no longer, in any way, bound by the limitations of nature, but are free to puruse our own ends in an environment free from natural consequences, a heavenly world solely of mind. However, unlike the pictured posthumous utopia of the dominant religions, this utopia would go to the survivors: those who could preserve their patterns long enough that they could forever save them from the continous onslaught of natural forces which tended towards entropy and thus the destruction of highly complex patterns, preferring diffuse apatternistic formations which tend toward equilibrium. Exact progress towards comprehensive models of either the human brain or low-level human biology is difficult to determine. Save for Jeff Hawkins, there are no notable people working toward the former goal, and none that I have ever heard of working toward the latter. Until humanity gets its priorities straight, realizes that AGI, or rather, “Friendly AI” is a moral imperative, and chooses the best route for pursuing this goal, it will remain elusive. The more humanity becomes focused towards achieving AGI as a goal, and focuses on realistic prerequisites, the sooner AGI will happen. I very much wish that humanity would focus its efforts on the creation of either a comprehensive theory of the human brain, or perfecting cellular biomodeling to the point that complex multicellular entites could be simulated. These two pursuits will, more than any other, bring us closer to Singularity. As to which one will win out in the end I am uncertain. At this time BlueGene is beating BlueBrain in terms of total computational power, and that is all I can say. The exact prerequisite structure for either task to fall into place as AGI is unknown and immense, so at this point, I think it’s really either discipline’s game. May the best discipline win. I just hope the public will eventually realize the need to support you both. The Singularity could occur within 5 years. It could occur within 50 years. Who’s to say? The sheer number of assumptions which are required to predict a timeframe are incomprehensibly immense, at least for any human. My ultimate point is that the Singularity isn’t anything to plan your life around: it will happen when it happens, and no sooner, or not at all, and we will all die. Enjoy yourselves for now, and know that the status quo, or thereabouts, is going to stick around, getting a little better all the time, until the final trigger is pulled and AGI is created. And while I may see that event happen with a head full of wetware, smirking at these remarks, until the prerequisite triggers are in place, to the general public all of you Singularitarians and transhumanists all sound like jackasses for talking about the transhuman age where we will live on as immortal transhuman entities. You can’t predict the trigger, I can’t predict the trigger, and until that happens your fantasies are merely science fiction. The birth of Friendly AGI will truly represent a day when the universe changes.

I agree with everything said here; however, this article seems to misunderstand the definitions and ideas of transhumanists and singularitarians.  Nothing in either philosophy necessitates foreknowledge of the future.  Nobody can predict the future.  People like kurzweil are just pointing out trends.  I think Kurzweil believes the Singularity is likely to happen in the first part of this century and only gives dates to spark discussion and spread the meme.  It seems that he has been quite successful.  The whole fact that this article was written is a prime example of how his ideas have shaped the discourse to some extent.

I think Kurzweil's ideas are a good meme to spread for this same reason.  It is getting people to think.  People typically only care about things that are within the short term, or at least their lifetimes.  Since the Singularity is actually likely to happen relatively soon, it is good to think about it.  Thinking and talking about it actually might speed it up because of the increased interest, funding, and research.

 
Whether or not the Singularity will be desirable is a whole other question that deserves some serious discussion.


I was speaking with my friend about this after a poker game. I lost 15 dollars, and he asked if I was disappointed. I told him, "No, because I only bet what I can afford to lose."  Then, whilst on the subject of the Singularity he said we should not gamble what we cannot afford to lose.

At first, I thought this was an interesting concern. I quickly rejected this though because we are gambling with our existence every single day.  In fact, the longer we are stuck on earth, the greater the chance that we will be anihilated quickly.  Ultimately there could very well be some heat death of the universe or multiverse, etc. Although, one way around that is the subjective extension of time to an incredibly vast amount of time - but that is getting off subject.  We will almost certainly never be able to find out about any of that stuff if we stop technological advancement and stay on Earth.

I don't know how it is possible to ban just AGI, you would have to ban all new technologies. I am opposed to that. Technology is our chance! We are gambling anyways, and certain to die if we don't...  so let's roll the dice!

Kurzweil is correct that we are in an unprecidented age. AGI is just one technology that will be needed to manage the singularity. Weather human biology is decoded and manipulated with precision in 40 years or not, it does not matter. In forty years we will have compelling regenerative medicine and perhaps the tools and machines to delay onset of death systematically.

We need utopian predictions to help move the science and the hard scientific investigation to new levels of achievement.

Reverse aging or the suspension of death is the key technology that will trigger a singularity.

If AGI is faked but helps millions of people, who cares? We are going to fake agi before we biologically create human intelligence in sillicon.

We are living in an age of exponential growth. Given the times we live in anything can happen, good or destructive.

I should preface this by saying that I'm not all that familiar with Kurzweil. What I do know of his thoughts has left a favorable impression (though his list of dietary supplements boggles the mind) despite the fact that I disagree with him on a number of points.

The article though, seems to ignore what I understood to be Kurzweil's basic, and fairly well known, premise. That he calculates based upon historic trends. He "assumes" that innovation will inevitably emerge, not out of a lack of understanding regarding the varied and unpredictable mechanisms of such innovation, but strictly upon the fact that historically innovation HAS always emerged in a timely and consistent manner. The transition to new paradigms has always allowed technological trends to continue upon the same curve. Yes, spontaneous human innovation is unpredictable and unreliable, but the totality of humanities spontaneous innovation has always been sufficient to continue the trend of technological advancement that Kurzweil has observed in the past. He, therefore, not unreasonably assumes that this will continue to be the case. One could assert that his sample size is too limited to make such a bold conclusion and then point to the vagaries of “spontaneous human innovation.” But to rant on about Kurzweil misunderstanding the nature of human technological advancement is obtuse.

Was Vinge wrong, too?