Thursday, December 18, 2014

The Approaching Technological Singularity

The Singularity:  “a technological singularity is a predicted point in the development of a civilization at which technological progress accelerates beyond the ability of present day humans to fully comprehend or predict.”

The singularity most under discussion these days will occur when artificial intelligence (“AI”) achieves the ability to mimic human consciousness.  The idea is all over the place.  News websites run articles about it, TV shows incorporate elements of it in their plots, and more or less serious publications like The New York Review of Books and Vanity Fair run big, almost scholarly stories about it.  Not to mention that granddaddy of cultural icons, the Terminator movies, which are all about machine intelligence run truly amok.  What, people wonder, will happen when machines outstrip us in intellectual ability?  What indeed. 

I would suggest that modern computers, the Internet and smart phones have already confused us sufficiently to fit the above definition of a technological singularity, but that’s just me.

The AI Debate

The advent of machine intelligence, in the form of primitive computers, came during World War II and immediately featured speculation about what might happen when these machines really get some wind in their sails.  Alan Turing, the father of the modern computer, was already thinking about it.  When will computers become able to mimic human intelligence?  He came up with a test that is still used today, Turing’s Test.  Human interrogators blind test a few people and one computer to see if the computer can fool them into thinking that it is one of the human test subjects.   They’re getting pretty close by this time. 

Part of the discussion is Moore’s Law, which hypothesized that the capabilities of computer chips will double every two or three years.  This is actually what has been happening for some time now, and the signs are that the progress will continue apace.  But for how long?  Will this tendency go on indefinitely?  If it does continue to grow at that pace AI will achieve capabilities that we can only guess at, and very likely this will happen in our lifetimes.   (Not mine, perhaps, but probably yours.) 

There is a very active debate in progress regarding this impending breakthrough.  Many talented scientists and tech geniuses are understandably fascinated by the prospect of machines that can think like people do.   The discussion is very heavy on “when,” and the “if” seems to be a given.  On one side are people who are very gung ho about the coming breakthrough in machine intelligence, the coming singularity.  Call them the Utopians ; they are also being referred to as “Singularitarians.”  On the other hand are the Cassandras, the nay sayers.   In the middle are many people who range from mere curiosity to a mild but active interest.  The curve is surprisingly flat; both extremes contain lots of people and the middle is only slightly more populous.  This is an area where opinions can be very, very strong.

The Singularitarians make amazing claims for the potential benefits of machines that can mimic the thought process of people.  Ray Kurzweill is a big time Utopian in this debate.  He claims that the Twenty-First Century alone will see 20,000 years of progress rolled into a mere hundred years.   Peter Diamandis, another Singularitarian, says that AI will achieve “exponential price-performance curves” and provide “plenty of clean water, food, and energy for all earthlings as well as decent educations and adequate health care.”  (In his book, “Abundance: The Future is Better than You Think.”)  Speculation about the coming changes and benefits are really wild, including the prediction that machine intelligence will marry with human intelligence and spread throughout the universe.  That seems like a stretch.  I’ll spare you a full reading of some of the famous techies that are waxing poetic about this new computer revolution.

There is a big push going on right now to bring about this singularity, to design and build computers that will mimic the human thought process with almost supernatural levels of power.  Many of our great minds are at work in the area.  There is actually a Singularity University in Silicon Valley.  It is located at the NASA Ames Research Center, no less, and it is funded by Google, Cisco Systems, Genentech, Nokia, and G.E.  Yes, I did say Nokia.  Their Nokia Research Center Cambridge at M.I.T. in Massachusetts is also working on the problem. 

The nay-sayers are a high powered bunch too.  They include such luminaries as Stephen Hawking, who has been all over the news in the last year warning that machine intelligence is coming, that it may not have our best interests at heart, and that it may indeed have the capacity and the inclination to do away with all of humanity.  That got my attention. 

Nick Bostrom of the Future of Humanity Institute at Oxford is worried too.  He is afraid that “human labor and intelligence will become obsolete.”  If we're lucky, the machines won’t bother to get rid of us all, but they may just allow us to live out in the woods somewhere as long as we are quiet and don’t make any trouble.  He points out, rightly I think, that it will be very hard to program goals into these new machines, goals that will not allow for any mischief.  It is, he says, “quite difficult to specify a goal of what we want in English, let alone computer code.”  He has a point there, doesn’t he?  I’d go further and suggest that if the machine were to actually think like a human being it could easily decide to disregard instructions in any case. 

Human Thinking and Behavior Are Messy

 The problem here is that the current discussion is about computers that will actually think with a naturalistic human thought process, ones that will be “fluent in the full scope of human experience” including “unusual but illustrative analogies and metaphors.”  (Mitch Kapor).  And the stated goal is to create such machines.   I believe that that is not only undesirable, but also impossible.  A machine intelligence will always be a machine. 

I think that the real danger here is that a true artificial intelligence could become a true machine entity of some new kind.  That it could become self-aware and that it could come to possess certain negative human characteristics, like ego, self-interest and the instinct for self-preservation.   Not to mention free will and autonomy.

This new machine entity would almost certainly not exhibit any of the sometimes messy intangibles of true human thinking.  Human consciousness includes components such as altruism, empathy, sentimentality, nostalgia, love, and the willingness to cooperate.  It is unlikely that a machine intelligence would develop these things on its own, and if they were programmed into the machine it could easily reject them out of growing self-interest or because they seemed ridiculous.

I wouldn't be surprised at all if a self-aware, self-interested, self-duplicating machine intelligence decided to just get rid of us as a bunch of ridiculous anachronisms.   What could we add to the new prosperity?  Humor?  Drama?  What could be more ridiculous to a machine than humor or drama?  And our life-support would be an expensive, unnecessary budget item. 

Machine intelligence will arrive as any number of separately constructed and programmed entities, and isn’t there a real element of danger in the fact that all of these machines will be able to communicate with each other and could choose to join forces in the name of self-interest?  That would be logical after all, and machines are nothing if not logical. 

So, I’m dubious about this whole thing.  I’m not going to get too nervous about it though, I’m sure that you’ll agree that other issues are making greater demands on our worrying time.  And a “Bengazi!!!” to you too.

Uncredited quotes in this post are from “Enthusiasts and Skeptics Debate Artificial Intelligence,” by Kurt Anderson, a recent article that appeared in Vanity Fair Magazine. 

Also of interest:  “AI May Doom the Human Race within a Century, Oxford Professor Says,” an interview with Nick Bostrom of the Oxford Future of Humanity Institute that appeared in August, 2014 on Huffington Post doc com. 

Also check out the movie:  “Colossus:  The Forbin Project.”  Computers such as those envisioned here are created and it all goes to hell faster than you can say “Jack Robinson.”  

No comments: