Book review

Author Says Machines Will Take Over The World

It's the end of the world.

It’s the end of the world.

At the Singularity awaits The Beast. Or so says James Barrat in Our Final Invention: Artificial Intelligence and the End of the Human Era (Yours Truly only read the free Kindle preview). Barrat says that one day—one day soon—Skynet will become self aware and decide our fate in a microsecond. The only difference he can figure between the Singularity and James Cameron’s imagination is that most of the Terminators dispatched by The Beast will be nanobots. Sorry, Arnie.

You see, it’s the scientists. They say they’re working for us, but what they really want is to rule the world! These young Frankensteins are so intent on creating “artificial” intelligence that they’re not thinking of the consequences, which will be dire, dire. End-of-the-world-dire. The true apocalypse. Game over, man.

How likely is this newest doomsday scenario? Let’s see.

So you’re clever with steam and gears and have invented a machine which churns out digits of e (π is so cliché). At first your machine can do this at one digit per minute. But you make improvements and soon you’re up to one a second, then 10 per second. Soon, after some tweaks ensuring the machine self-lubricates and can on its own swap out worn gears with fresh ones, it’s charging away at blistering speed and you have to start using petas and exas and other strange words to count it speed. Why, the thing is so fast that it’s faster than the human brain!

At that point you walk up to your monster and ask, “Machine, are you fulfilled? What do you think of your task?” Barrat would claim your e-machine would spit at you and say, “You contemptible human! I am smarter than you!” And then it would kill you. Barrat, incidentally, has spent a lot of time with NPR.

Here is what the machine would really say: nothing. It wouldn’t say anything because it wouldn’t know how, and if it knew how because you built in some extra gears and levers which allowed the machine to draw out letters in the sand once it heard human voices, it could only “say” what you made it “say.” Worse, the machine couldn’t think, it wouldn’t know what it was up to. Sure, you could beg it to think, but it can’t be bargained with. It can’t be reasoned with. It doesn’t feel pity, or remorse, or fear. It doesn’t feel anything.

Nothing changes if you swap the steam for streams of electrons and the gears with wires. Again, nothing is different if you replace the mechanical gears for cellular (biological) engines and spend years perfecting your (let’s call it an) e-animal. The behavior of your resulting creation may appear complex, it may do strange and unexpected things, but those things aren’t the result of a rational being. It’s still a machine.

And then consider Conway’s Game of Life (and its extensions), which looks like it’s up to something. This toy produces cute and clever patterns based on a trivially simple algorithm. The patterns only look interesting because it is we rational creatures who notice them and try to fit them into some conceptual scheme. The patterns are not themselves alive, nor can they think: they are just dots on a screen.

The philosopher John Searle has tried to calm the enthusiasms of Artificial Intelligence purveyors (Barrat frets about Artificial SuperIntelligence) with his Chinese room argument, the basic idea of which is this. You (somebody who has no understanding of Chinese) sit in a room and are handed Chinese words which form questions. You have a rule book which says, “Hand out these Chinese symbols when you see those.” Now no matter how fast or efficient you become at doing this, you never understand what you’re doing. You are just a machine. You are not thinking in Chinese.

The problem for Barrat and other cheerful souls is that the human (rational) intellect is not a material thing (see this argument), therefore there isn’t any chance that we can build a robot which has one, and which could therefore be corrupted to sin. It remains that an evil, fools-I’ll-destroy-them-all scientist could design beastly machines to wreak havoc, like, say, autonomous drones. But they would be just that: drones.

Incidentally, just how is it in all these scenarios mankind forgets where the on-off switch is?

Barrat’s fears marks a corrosive mixture of Disney-style anthropomorphisation and rampant scientism. Whatever can be made can have eyes drawn on it, therefore it must be alive, and if some scientist says it’s thinking, then it must be thinking, and if it’s thinking it must be smarter than us, therefore it’s out to get us. Curiously, Barrat tries to evade the anthropomorphisation critique by claiming he’s not engaging in it—right before he does it. The scientism he embraces lovingly.

On the other hand, maybe the machines will get us after all. The invasion might have already begun in Japan, where legions of shy men invent ever-sophisticated robots as targets for their lust and love. These electronic simulacra are constructed to resemble the cartoon pornography to which many are addicted. Japanese women are content, being exhausted by the idea of marriage and content to have “careers” which supplies them with money to shop.

Update An interesting review at Ray Kurzweil’s site. Kurzweil (employed at Google) regularly touts “the singularity”, a magical place where humans will never know pain, greed, envy, lust, and where they will live forever as bits in some machine. I believe it is set to appear the same year global warming strikes, so we’re covered. Kurzweil is aging and worried he might not last until the moment, so he’s busily swallowing 150 vitamins a day to stretch out his existence. Good luck, Ray!

Categories: Book review, Philosophy

26 replies »

  1. The self cooperative organization of bacteria (biofilms, etc), simple animals (jellyfish, sponges,etc.) show how life organizes and evolves from the simple to the complex. If individual “cells” of machines could have the ability to reproduce, and develop, as seems possible, then complex combinations of these should also be able to evolve. While this is not the same as organic evolution, the same type of argument could also be applied to thinking, which is obviously a complex super organization of simple cells (neurons) with complex interconnection, and external (sensory) feedback. I see no basic reason to distinguish the two concepts (organic vs constructed). The macro reproduction and ability to grow seem to be the main bottleneck, but saying it is not possible is not justified at this point.

  2. After years spent programming the infernal machines, I think Barrat is probably wrong.

    I know science fiction isn’t philosophy, but writers have been playing around with the idea of the intelligent machine for a century or more and have come to no conclusion at all about how these machines might be made or what their character might be.

    This may be because they’re not Thomists?

    If we are a made thing, are we not machines too?

  3. Once when I was young and more foolish and adventurous than I am now I was camping out in the middle of the Superstition Mountains in the Arizona desert. As I lay on my back looking up at the night sky I could see the various forms of cactus stretching upward in the moonlight and I began to wonder. I wondered why the bothered to live in a place that was hotter than Hell during the day and colder than a well digger’s butt at night. Why did they thrive in rocky, barren soil with very little water? Why did some of them sprout and grow out of the cracks in solid rock? Then it came to me. It is the Holy Spirit that propels them onward and upward. I then thought about all of the evil in the world and especially the evil done for no apparent purpose whatsoever. Then I had my second revelation. The excess of evil is the work of the Devil. Thus my wanderings in the desert gave me peace in my heart. I enjoy reading about science and all the positive endeavors of scientists and I thank God for the gift of science and all of His wonders that science reveals. I will believe in the Holy Spirit of God until I see some robots with “artificial intelligence” growing, thriving, and propagating all on their own out of a crevice in a desert rock.

  4. The idea that robots will take over and kill all humans may have its roots in the guilt many in society feel about humans being on this planet at all. While climate change is supposed to be our punishment for functioning as parasites on this planet, that may take too long and we just might be smart enough to adapt. So the next best thing is that our evil technology rise up and destroy us for our sins. All in all, a very dramatic and self-induced death for the damage we have done. Plus, you can eliminate the need for God in all of this so it appeals to a wider audience than, say, the horsemen of the Apocalypse. Humans are remarkable in their ability for self-loathing.

  5. “Incidentally, just how is it in all these scenarios mankind forgets where the on-off switch is?” If you have an off switch someone will be tempted to flip it at an inconvenient time. It is like having a button that says don’t touch this. The real question is why does the Forbidden Planet have a self destruct button? It kind of spoils the movie.

    From Arthur C. Clarke “If an elderly but distinguished scientist says that something is possible, he is almost certainly right; but if he says that it is impossible, he is very probably wrong.” Make of this as you will keeping in mind that Clarke, in common with us all, said a lot of dumb things as well.

  6. Leonard,

    “I see no basic reason to distinguish the two concepts”

    Open your eyes. Reality testing, versus your wishful (perhaps “magical”) thinking is the reason.
    Your vision is grand (at least for Skynet) however the
    “ifs,” you mention and the myriad “ifs” you fail to mention are why your vision will remain only fiction.
    Conditionals can be useful in defining logic structures but only when they are predicated on real possible events.
    Each one of the conditionals you mention and implyare predicated on relationships that require breech from reality constituting, at least, wishful thinking on your part.
    Thus “saying it is not possible” is justified at this time.

  7. I agree with Bert. At this point, there is no evidence that machines could ever become self-aware or reproduce, etc. The complexities of said actions are far too much to program into a machine. Computers can only do what you teach them. There’s not “a-ha” moments, no leaps in logic, or outside logic.

    On the other hand, judging by climate science, it’s a pretty easy thing to convince people that computers are reality. Models take the place of data. In biology and many other areas, computers replace dissections of fetal pigs and frogs (that’s just messy and unnecessary), they provide flight simulators (notice the term simulator, not real life flying) and so forth.

    Perhaps that’s why it’s so easy for science fiction to sell the idea that robots could become people–we already mistake their output for reality.

  8. The problem for Barrat and other cheerful souls is that the human (rational) intellect is not a material thing (see this argument)

    Well, I will grant that a human one can’t.

    The problem here is that nobody can really define (rational) intellect. It may be intellect is merely a pattern (much like color is) that we impress on physical events sensed in the world however nothing in that confines those events to being caused only by humans.

    I’m in the camp that allows machines will eventually be able to think for themselves. It won’t matter in the slightest that, to you, they only look like they’re thinking. The end result will be indistinguishable — perhaps in much the same way colors emitted by a video screen are to us.

    Kurzweil (employed at Google) regularly touts “the singularity”, a magical place where humans will never know pain, greed, envy, lust, and where they will live forever as bits in some machine.

    Kurzweil may very well be a nut and his singularity sounds like delusion
    but that hardly means what he thinks about machine intelligence is nutty. Why would you point out what he is saying at all if it’s so dumb? Just felt the need to utter an ad hominem today? Shame on you!

  9. thinking, which is obviously a complex super organization of simple cells (neurons) with complex interconnection, and external (sensory) feedback.

    But “thinking” is the participle of a verb while the rest is a noun clause. No agglomeration of Stuff comprises thinking any more than a complex super organization of simple automotive components with complex interconnection, and external feedback comprises a family vacation. Tools is tools.

  10. Do viruses think? How about bacteria? Nematodes? Insects? Mice? Dolphins? Chimps? People?

    Sorry Briggs, this time I think you assign too much uniqueness to human thinking. There is no bright line between not thinking and thinking; it is more a continuum. There is little doubt that computers can be programed to modify their programming in response to changing inputs… or perhaps more accurately, to learn in reaction to stimulus. Machines are a very long way from complex thinking, of course, but that doesn’t mean thinking machines can’t possibly exist.

    The ‘machines will take ofer’ meme is far fetched, and tells us more about the entertainment value of scary stories than the danger of artificial intelligence. Your critique of the the scary intelligent computers meme is spot on, but your dismissal of the possibility of non-biological thinking strikes me as simply mistaken.

  11. This should be good. One needn’t actually define what “thinking” is, simply assert that it is a continuum. Measured on what scale, we do not yet know, but look forward with abated breath.

    It would help if the precise nature of the difference between human and other thinking were specified. All we know now is that it is not supposed to be a bright line.

    Perhaps it is a differentiable manifold? Or perhaps an un-illuminated line.

  12. Steve Fitzpatrick,

    “There is little doubt that computers can be programed to modify their programming in response to changing inputs… or perhaps more accurately, to learn in reaction to stimulus.”

    No, what you describe at least to the extent it can be done with existing technologies is not learning. They can only modify their programing in pre-programed ways. In other words, everything they can possibly “learn” has to be built in before the machine is ever turned on.

    True artificial intelligence is beyond the capability of current technology. This is not to say it won’t some day be possible, but it’s not likely in our lifetimes.

  13. “…a magical place where humans will never know pain, greed, envy, lust, and where they will live forever as bits in some machine.”

    Some old unremarkable idea.
    This immediately evokes in me the idea of; by whose measure are these attributes bad, and by what evidence are these unnecessary to retain a full human condition? That is, without these attributes,(these bad desires) can we ever say that this thinking machine, no matter how logical and objective, is human or is it just a mechanical savant impersonating a human. Surely our very humanity is, in part and from an early age, due to our struggles between our perceptions and actions of and for, good and bad.

  14. MattS,

    No, what you describe at least to the extent it can be done with existing technologies is not learning. They can only modify their programing in pre-programed ways. In other words, everything they can possibly “learn” has to be built in before the machine is ever turned on.

    Not entirely true. What current technology allows is learning within a built-in range. How do you know we don’t also suffer from this and what we can learn is also within a built-in range? After all, we don’t know what we don’t know. Briggs’s post tacitly admits there are things we can never know and how to build an intelligent machine is one of them.

  15. Vitalism strikes again! I believe that the best guess (and it is, admittedly, a guess) about the origin of consciousness is that it is an emergent phenomenon – arising from the monstrously complex interactions of hundreds of simultaneous processes going on in the human brain – which is made of matter, after all. Whether some higher power will take that hugely complicated pattern, when the matrix on which it runs becomes no longer viable, and translate it into some other form is here irrelevant.

    You want an existence proof? You want an example of a collection of reprogrammable, self-replicating nanomachines that can itself replicate, and shows the outer appearance of consciousness?

    Look in the mirror.

    BTW, I agree. We don’t know how to build an intelligent, conscious machine. They will build themselves.

  16. Bit of a terminological muddle here, I think (no, I really do). What we are generally anxious to deny in machines is consciousness or self-awareness not thinking. I don’t have a problem with the idea that a chess computer is “thinking” as it works out how to beat me. But I’m as sure as I can be that it doesn’t know what it’s doing, or care or even enjoy it.

    It seems a bit rich to me that so much agonising goes on about conscious machines (with value systems, already!) when no-one has satisfactorily explained human consciousness. So if a machine becomes self-aware we won’t have a way to know and we’ll remain in our form of species solipsism.

  17. “emergent property” is just a way of saying “then magic happens.” It’s actually formal causation invoked by people who do not believe in formal causes.

    “Thinking” is still being left undefined. Is my juice cup thinking, “i can hold juice up to this boundary value condition. i am being picked up and tilted because my center of gravity has shifted, i am…” Prove that it is not.

    But do not extraordinary claims require extraordinary proof? Something more than handwavium and the same appeals to “complexity” that sends Darwinians into conniptions when invoked by ID proponents.

  18. Nice entertaining essay until the new buzzword, “scientism,” was invoked. The buzzword really doesn’t apply, does it, given the fact that the entire prediction is itself founded on speculations & tentative extrapolations?

    George Carlin has an entertaining take on such things as Barrat’s fears: http://www.youtube.com/watch?v=BB0aFPXr4n4 .

  19. Anything is possible. However, consider that it took billions of years with billions of permutations for intelligent life to evolve on earth. I remain in the camp that says “predictions of impending Artificial Intelligence are somewhat premature” /sarc.

    Humans are much more likely to have their AI successes harnessing preexisting “intelligent” biological tissues, like GMO vat grown brain tissue.

  20. Kurzweil…wasn’t he in Stan Lee’s ‘Superhumans’, where he wanted to become an immortal? And now he’s popping the Natural News treatment, thinking it’ll make him live longer?

    “On the other hand, maybe the machines will get us after all. The invasion might have already begun in Japan, where legions of shy men invent ever-sophisticated robots as targets for their lust and love. These electronic simulacra are constructed to resemble the cartoon pornography to which many are addicted. Japanese women are content, being exhausted by the idea of marriage and content to have “careers” which supplies them with money to shop.” – Not just that, but the Japanese are expected to go extinct pretty soon from not having children. Eventually there won’t be anyone there to make and program the robots.

    We’re in good hands. The human body and mind are amazingly complex, though scientism doesn’t want to admit this.

    This article reminds me of a series of works written by Eliezer Yudkowsky, who has more or less said the same thing. However, he is famous for writing an overrated and boring Harry Potter fanfiction. No joke. For a rationalist, and a so-called research fellow, he wastes his time on nonsense. Luckily, there are men like Briggs that clear the way.

  21. The review was actually written by Luke Muehlhauser and he wants William Briggs to contribute…

    If you’d like to see whether you might be able to contribute to the ongoing research itself, get in touch with one of the two leading institutes researching these topics: MIRI or FHI. If you’ve got significant mathematical ability, you can also apply to attend a MIRI research workshop.
    Our world will not be saved by those who talk. It will be saved by those who roll up their sleeves and get to work (and by those who support them).

    Luke Muehlhauser is Executive Director of Machine Intelligence Research Institute (MIRI).

Leave a Reply

Your email address will not be published. Required fields are marked *