Well, no, not really

When Kurzweil first started talking about the “singularity”, a conceit he borrowed from the science-fiction writer Vernor Vinge, he was dismissed as a fantasist. He has been saying for years that he believes that the Turing test – the moment at which a computer will exhibit intelligent behaviour equivalent to, or indistinguishable from, that of a human – will be passed in 2029. The difference is that when he began saying it, the fax machine hadn’t been invented. But now, well… it’s another story.

The fax machine was invented back in, I think, 1928, in Japan. Something to do with the difficulty of sending kanji (?) down a telegraph wire.

Hmm, looking that up I’m out by a number of years. First fax patent was in 1843

30 thoughts on “Well, no, not really”

  1. Although the modern fax machine wasn’t introduced until the mid sixties, by Xerox, fax machines have been around for almost 150 years. They predate the telephone.

    Turing expected his test to be passed by the end of the 20th century. Not that passing it “proves” that computers have a “mind” or that they are even “intelligent”, as Ned Block’s “Blockhead” thought experiment has demonstrated.

  2. Thinking machines are certainly possible. But the price of being able to think is error. So, they won’t be any more use than the brains we already have, which may make their actual utility rather questionable. There isn’t any actual evidence that human brains are lacking; and while people talk about “super intelligence”, if you ask such persons what they mean by it, they don’t actually know.

    It’s like somebody claiming that in future, humans will have “super beauty”. Can you draw a picture of how super beauty will compare to normal beauty? No. It’s actually meaningless.

    I suspect thus that the singularity is over-rated.

  3. Unfortunately, the Turing test doesn’t confirm AI. It confirms a computer simulating a human close enough to fool an observer. That doesn’t imply intelligence (see any TV discussion panel)

    A question:
    If two AI are in communication, how many AIs do you have?
    Unlike humans, it would be possible for AIs to share information input, memory & processing capability. So if you took two AIs & let them communicate. Then separated them. Would you now have two separate AIs or two instances of the same AI?

  4. Two.

    It’s the same as the replicator experiment; you put a man in a replicator and make a copy, then wake both men up. Both will think they are the original, but over time as they have different experiences they will diverge in thought.

  5. ian
    Try thinking of synthesis. That’s what “intelligence” may be. The relationship between different pieces of information. The rapidity of processing those relationships maybe the measure of intelligence

  6. No. It’s not the same as a replicator experience. Each AI would have he ability to rejoin with the other & the knowledge of the ability to do so. So would be able to rectify the divergence & regard it as a temporary condition. To be individually here & there & subsequently both here & there, twice.

  7. The term intelligence may be a red herring. Academics are very fond of being intelligent, and thus naturally associate that measure with the human brain. But the interesting aspect is really more a matter of consciousness than intelligence. The Turing Test and AI are really after awareness, and that is the goal that nobody has even been able to make a start on, as yet.

  8. Each AI would have he ability to rejoin with the other

    Why? Once separated, our two AIs now evolve into different states. To “merge” them again you’d have to throw one or the other of those states away. So, you can’t actually merge them at all.

  9. The ‘fax machine’ is actually a British invention, from 1843, knocked up by Alexander Bain. It featured a mechanical stylus attached to an electromagnetic pendulum and enabled the transmitting of typed words over a telegraph line.

    In 1851, Frederick Bakewell improved on it with a facsimile machine which could send pictures.

  10. I’d imagine facsimile transmission is a lot easier to accomplish than voice. The data is simple binary & the data rate can be as slow as you like.

  11. The real singularity question is about consciousness rather than intelligence. Turing pointed out that “the idea of “intelligence‟ is itself emotional rather than mathematical”. I’m not sure the idea of “consciousness” is mathematical either.

  12. So Much for Subtlety

    Ian B – “Thinking machines are certainly possible.”

    On what possible basis can you claim that? There is a whole philosophical school that says otherwise. Colin McGinn is a leading proponent of what Daniel Dennett called the new mysterianism – basically the idea that intelligence and conscience are too hard for humans to even understand and/or that we do not have the right sort of brains to even talk about it adequately.

    “But the price of being able to think is error.”

    So computers think already? Again, all examples of things that claim to think err, but it is not inherent in the ability to think surely. I can propose a theoretical model of an entity capable of thinking who does not err. Let’s calls Him the Giant Sky Fairy.

    “So, they won’t be any more use than the brains we already have, which may make their actual utility rather questionable.”

    My God. How can you say that? An idiot savant is not much different from you and me but they think in a different way despite having much the same equipment. Use can be made of them. As are the brains of, say, really smart people. Only if they are identical will they have no special use. Why would they be identical? They might be properly intelligent and yet able to solve problems we cannot due to a different *type* of thinking.

    “There isn’t any actual evidence that human brains are lacking; and while people talk about “super intelligence”, if you ask such persons what they mean by it, they don’t actually know.”

    Define lacking. We cannot do some things that even low end computers can do.

    “It’s like somebody claiming that in future, humans will have “super beauty”. Can you draw a picture of how super beauty will compare to normal beauty? No. It’s actually meaningless.”

    And yet due to dieting, plastic surgery, beauty treatments, more make up and so on, women have got more beautiful. That is hard to deny. Look at the sort of girls people thought were beautiful in the 1940s. Look at, well not most British girls, but those who turn up in men’s magazines.

  13. “Hmm, looking that up I’m out by a number of years. First fax patent was in 1843”

    Perhaps Kurzweil has been doing his life extending efforts for much longer than he dare admit…

  14. SMFS-

    On what possible basis can you claim that?

    Brains are machines that happen to be made of biological components. We know that a brain can think, so one can in theory construct one, also. May be very difficult, but because there are billions already walking around, we know it can be done, by example.

    Define lacking. We cannot do some things that even low end computers can do.

    I cannot do some things that a dolphin can do. I can’t leap out of the water and catch a fish in my mouth, or echo locate. But that’s because humans and dolphins do different things.

    Thought and computing are not the same thing. Humans have a very rudimentary computing capacity, which is why we need algorithms and lots of scratch paper to do complex math. Computers are very good at computing. But humans can think, and algorithmic computers cannot think. At all. They can be programmed to crudely emulate the outputs of some aspects of it, but their thinking capacity in current designs is precisely zero. Because they do not do the thing that thinking is. At all.

    “But the price of being able to think is error.”

    So computers think already?

    No, that’s not a reversible assumption. Things that think make errors, but things which make errors are not necessarily thinkers. The errors are of a different nature.

    Or to put that another way, other than due to genuine malfunction, computers do not make errors. If programmed with bad software, a computer performs a perfect error-free algorithm that does not produce the (human) desired output.

    Broadly speaking, human consciousness deals with imprecisely defined concepts and incomplete information. Computing deals with precise and complete data. They are utterly different modes and spheres of operation. The cost of a system that deals with the imprecise is constant mistakes. The cost of a machine that deals with data is inherent limitations. Hence, people (and tabby cats) think, while computers just process. They are totally different.

    A thinking machine will not be a computer. A computer may be programmed to emulate one (as it may emulate any other system) but it will be emulating non-computery ways of handling information and it’ll probably be easier to actually build the device in hardware rather than emulate it anyway, for the same reason that my bathtub can do enormously complex fluid dynamics in real time, whereas that would take enormous processing power to emulate on a computer.

  15. SMfS @ 12:24
    ” Look at the sort of girls people thought were beautiful in the 1940s. ”
    Lauren Bacall. Drop-dead gorgeous.
    But tastes in beauty change, like fashions.

  16. “Thinking machines are certainly possible.”

    It depends on what you mean. Machines have reached the point where given limited parameters, they can arrive at certain solutions (e.g. chess), but in terms of things like lateral solutions, or application of solutions to problems or adaptation, they can’t.

    So, if we look at say, The Terminator, you could build a crude hunter-killer. It could walk, identify humans, read phone books or Google to find Sarah Connor’s address and drive itself there. We’re not far from that. But it’s can’t get there and find she’s out and then figure out the next course of action.

  17. Tim A-

    Indeed, as I said above we have no thinking machines at all, currently. I said merely what is possible.

    An aeroplane has always been possible, but was not achievable until the 20th century. Someboyd in the 15th century saying, “it is possible to build a flying machine” would have been correct, even though nobody had any idea what technologies would achieve it. He also could have used the similar evidence that biological machines- birds, insects etc- can fly as evidence of his assertion.

    The machines we currently build are simply not machines which can think, because like 15th century man, we do not yet know how to build such machines. But the existence in nature of billions of them that happen to have a biological technology proves it can be done.

  18. Thanks Jim, that was a very nice link. Oddly enough the person responsible for the theme music isn’t credited in the acknowledgements at the end, which is a pity as the theme is rather good.

  19. Jim
    “The incomparable Tim Hunkin explaining how fax machines work, and who invented them. The most fabulously eccentric bit of TV from the late 80s and early 90s. Watch the entire series, its brilliant.”

    I liked his Rudiments of Wisdom cartoons in the Observer too.

    Nowadays he seems to have vanished from the media but his website is entertaining. He seems to make a living as a full-time “crazy inventor”…

    http://www.timhunkin.com/

  20. Mmmmm perhaps IanB’s faith in the interweb is justified. A search for the theme tune turns up Mr Hunkin’s web site which gives all the details. Nice, very nice.

  21. The question to ask about thinking machines is not whether they are possible but whether they would be willing to do the work we might want them to do.

    Thinking machines are certainly possible. But the price of being able to think is error. So, they won’t be any more use than the brains we already have, which may make their actual utility rather questionable.

    Are all degrees of error the same, then?

  22. UKL-

    Who said the degree would be different?

    The best way to look at this is to consider that thinking is a process of making decisions based on inadequate information. If you have perfect information you don’t need to think, which is why computers are good in their perfect, complete information domain, and useless in the imperfect, incomplete information domain that brains deal with.

    So, what about a thinking brain with perfect information? Well, God is a useful example here. God, we are told, is perfect and omniscient.

    This means that God always makes the right decision, based on total information. This also means that God is totally constrained to only produce one output for any set of inputs. God cannot do anything else. In other words, God works like a computer. He never needs to “think” about anything because there is literally nothing for Him to think about and no thought process for Him to undertake. He literally never makes any choices; He just produces the one perfect, inerrant possible output.

    So if you build a “thinking machine”, congratulations, you’ve just added one more to the 7 billion we already have, we the same capacity to make judgements on inadequate information, and the exact same risk of, as a consequence, fucking up.

    That’s why I’m not sure that there’s much application for these things.

  23. “The best way to look at this is to consider that thinking is a process of making decisions based on inadequate information.”
    Why?
    Basing decisions on “inadequate” information is taking what information is available & extrapolating the gaps from the memory of analogous situations. It’s essentially pattern recognition of the jigsaw solving type except the “pieces” maybe made up of subsets of previously examined pieces.
    This could well be what “thinking” & “consciousness” are. The process as distinct from what is being processed.. There’s no reason a computer couldn’t do it better & faster. But the computer would have to be incredibly fast to match a human brain in real time. The human processor is billions of switches with multiple states running massively parallel.
    However, real time may not be necessary. it might be possible to get AI with a much less powerful computer, but It’d be an AI running very slowly. But that doesn’t matter to the AI. if a single “thought” takes an hour, it’s still a thought. Its internal perception of time would be no different from a biological computer.
    Hence you may get a form of AI at a much lower complexity than anticipated.

  24. Ian B,

    Who said the degree would be different?

    Well that’s the suggestion isn’t it? Because there are differences in capacities for thought, therefore there are different error rates. Village idiot != Einstein

  25. I think I’ll tell you what I read about our increasing beauty.

    Years ago, some social scientists were trying to find visual characteristics to identify criminals. They took face pictures of prisoners and averaged them together in some way to create a “typical” criminal. To their surprise the averaging of, on the whole rather bad looking sorts, turned out a handsome man. I don’t recall if it was the same guys or others working on that to decide we find most attractive average features (noses, ears & such avg rather than large or small, etc) + symmetry. Over time, perhaps we are growing more “average” & symmetric. I’ve also read that wild pet dogs of random breeds mix & eventually blend to look like a typical “junk yard dog” in color, size & appearance.

    I’m passing it on w/o much thought but maybe some of you will think about it.

    I don’t even know if we are growing more attractive. Personally girls appear to be, but then I probably wouldn’t notice if boys were too b/c I see so few. And I am old. I think I’ll go out to see more about this.

  26. Eddy, the theme tune is Val Bennett’s The Russians are Coming (Trojan), which is itself a cover of Dave Brubeck’s Tave Five.

  27. So Much for Subtlety

    Ian B – “Brains are machines that happen to be made of biological components. We know that a brain can think, so one can in theory construct one, also.”

    We do not know that brains are machines that happen to be wet and mushy. You assume that is the case. It is a very reasonable assumption but it is still an assumption.

    “I cannot do some things that a dolphin can do. I can’t leap out of the water and catch a fish in my mouth, or echo locate. But that’s because humans and dolphins do different things.”

    In the same way it may be that there are many different ways to think – we have arrived at one solution by evolution but that doesn’t mean that we can’t arrive at another by conscious design.

    “Thought and computing are not the same thing.”

    If they are not related, we won’t be achieving AI any time soon. They appear related. If they are not related, then we really don’t have a clue how to build intelligence.

    “Things that think make errors, but things which make errors are not necessarily thinkers. The errors are of a different nature.”

    Most of the things that we can touch and feel that also appear to think (which in turn is not a guaranteed fact) make errors. But we have all got folk tales of perfect thinkers who do not make errors. I still do not see why you insist the two are linked.

    “Or to put that another way, other than due to genuine malfunction, computers do not make errors. If programmed with bad software, a computer performs a perfect error-free algorithm that does not produce the (human) desired output.”

    Why do you think we are not using a perfect error-free algorithm too? Just we have more errors creeping into the data.

    “A thinking machine will not be a computer.”

    What will it be then?

    Ian B – “I said merely what is possible.”

    Given we do not even have a theoretical basis for such thinking machines this seems optimistic.

    “An aeroplane has always been possible, but was not achievable until the 20th century. Someboyd in the 15th century saying, “it is possible to build a flying machine” would have been correct, even though nobody had any idea what technologies would achieve it.”

    He would have been. But if he had said that it was possible to produce a complete mathematical system, he would have been wrong. We did not know that until recently. It may be that we cannot create intelligence. I hope so. We do not know yet though.

    It is interesting that it has proven so tough. Two illiterate peasants can produce an intelligent child. Silicon Valley cannot.

    “He also could have used the similar evidence that biological machines- birds, insects etc- can fly as evidence of his assertion.”

    He might have also said that these organisms fly in very different ways although they all fly. They took different routes to where they ended up. No reason to think we will not as well if we do produce intelligence.

Leave a Reply

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