back to article Stephen Hawking: The creation of true AI could be the 'greatest event in human history'

Pioneering physicist Stephen Hawking has said the creation of general artificial intelligence systems may be the "greatest event in human history" – but, then again, it could also destroy us. In an op-ed in UK newspaper The Independent, the physicist said IBM's Jeopardy!-busting Watson machine, Google Now, Siri, self-driving …

COMMENTS

This topic is closed for new posts.
      1. rcmattyw

        Re: I just wish....

        I imagine the physical structure of the environment would determine to a large extent how an intelligence would evolve. That would be why humans think in a certain similar way, although there are variations for various different disorders ADD, aspergers etc. Each having its advantages and disadvantages.

        Now imagine that consciousness evolving as you describe from within the frame work of a machine rather than our old brains. I imagine the resultant intelligence would be so far removed from our own as to be unrecognizable. We could create it without even knowing we had. Likewise, how is it possible to impose ethics developed by a completely incompatible intelligence (our own) onto the newly created artificial intelligence? It will likely be nothing like our own, so any ethics will be completely different to our own and arrived at in a different way.

        Even if we can recognize that a conscious being has been created, who are we to judge its sanity or insanity. It has evolved in a completely different way. I suspect anything which evolves will equate to madness in our eyes, but also seem to be completely brilliant at whichever conclusions and approaches it takes to a problem due to being so different to our own. But with a being so different, how are we to even communicate in the first place?

    1. Michael Wojcik Silver badge

      Re: I just wish....

      'ordinary' AI had reached a sufficient level of development that OCR would (a) actually recognize what's scanned and not produce gibberish

      It has, actually. You could make an OCR system that's much more reliable than the typical out-of-the-box matrix-plus-peephole-context approach using well-understood tech available today. You could even do it with free software: UIMA (for the processing framework) plus OpenNLP (for MEMM-based decoding), some glue code of your own, and a human-supervised kernel to start the training.

      People are building those sorts of systems for problem domains where they're profitable, such as scanning medical records.

      (b) provided me with a grammar checker that's a tad more sophisticated than suggesting that 'which' needs a comma before it or if no comma then 'that' should be used.

      There, I'm afraid, you're out of luck. Even expert human judges disagree widely on the matters that these so-called "grammar checkers" weigh in on. Grammar checkers are snake oil and a very poor substitute for learning to write well the old fashioned way: reading a lot, writing a lot, and getting a lot of feedback on your writing.

      (I won't mention the fact that "grammar checkers" actually evaluate text against a (suspect) set of heuristics for usage and mechanics, and have little or nothing to do with grammar.)

      That said, there are commercial usage checkers that are far more sophisticated than the crap that's built into word processors, too. I've seen demonstrations of them at conferences such as Computers and Writing. I wouldn't recommend anyone use them (because ugh, what a horrible idea), but they do exist.

  1. Lars Silver badge
    WTF?

    Oh dear

    With the knowledge of human history, Stephen Hawking, don't you think, after all, that the greatest threat to us is us, the education we should provide to our children but fail to deliver. Remember we invented the stone axe, and perhaps that was the biggest event in human history. We survived that too. It's not the tool it's all about how to use it. But perhaps your nightmare as mine, is when the stock exchange is run by AI.

    Oh well, still even if AI has been a very lucrative business for a very long time for the snake oil speakers at otherwise fairly honest occasions. Mostly the same AI con priests year after year. Perhaps the truth is right there in A as artificial and I as in intelligence. Quite a cake to program and run with a computer but how are we so damned stupid that we let the AI cons shit in our faces. I am sure we will have, eventually, artificial hearts running some fine program, perhaps a nice artificial penis, wi-fi perhaps. But a artificial brain, why the hell. Computers will become faster, smaller and so forth. But the AI cons are really hemorroides in the arse of IT. Don't pay those guys to have their speech, they will never deliver anything but air. And what the hell is it with you Google, have you become religious or something when you paint "arse" red. But ass bum but anus anal bottom booty asshole hole buttock is OK with your religion and then again not arsehole.

    Seriously Stephen, while you depend now on computers do you seriously believe your intelligence could be put in a box.

    1. Fink-Nottle

      Re: Oh dear

      > Remember we invented the stone axe, and perhaps that was the biggest event in human history. We survived that too. It's not the tool it's all about how to use it.

      There's a possibility that it was the Neanderthals who invented the axe, and sharing that invention led to their ultimate extinction.

      http://www.livescience.com/38821-neanderthal-bone-tool-discovered.html

      1. Lars Silver badge
        Happy

        Re: Oh dear

        @Fink-Nottle, from that article:

        "There are sophisticated bone tools that are even older in Africa, for instance," McPherron said. "Neanderthals were, however, the first in Europe to make specialized bone tools.".

        Apparently the Neanderthals reached Europe before "us" so there is indeed a logical link here to AI as for logic. Lots of words.

      2. donaldinks
        Mushroom

        Re: Oh dear

        "There's a possibility that it was the Neanderthals who invented the axe, and sharing that invention led to their ultimate extinction."

        *****************************************************************

        "At least one-fifth of the Neanderthal genome may lurk within modern humans, influencing the skin, hair and diseases people have today, researchers say."

        http://www.livescience.com/42933-humans-carry-20-percent-neanderthal-genes.html

  2. Anonymous Coward
    Anonymous Coward

    AI doesn't really have to be that advanced

    in order to unleash hell upon humanity. Any mediocre artificial entity will do just fine with the help of some of our fellow humans.

    1. Don Jefe

      Re: AI doesn't really have to be that advanced

      What Humans consider to be a highly intelligent member of their species is wrong about 2/3 of the time. A computer could probably equal, or exceed, that one day. But it still won't be able to look at itself in the mirror and say 'I wasn't wrong' and be right. Ha! Suck it computer.

    2. Anonymous Coward
      Anonymous Coward

      Re: AI doesn't really have to be that advanced

      I will speak it's name: The Corporate Workflow Management System - turning thinking human beings into fleshy effectors running in a booby-trapped rat-maze for The Gulfs pleasure!

  3. PhilipN

    Motivation

    Just because something is intelligent does not mean it necessarily wants to do anything

    1. Francis Boyle

      Re: Motivation

      Unfortunately what it wants to do will be determined by its creators and in ways that they don't even begin to understand (cf. children). And that's reason to be afraid.

    2. Destroy All Monsters Silver badge
      Terminator

      Re: Motivation

      Just because something is intelligent does not mean it necessarily wants to do anything

      Picture AIs streaming endless archives of football games to their data centers while they stroke their chips&beer perceptrons.

  4. Neil Barnes Silver badge
    Terminator

    I was noodling on the idea of AI a few days ago

    And came to the conclusion that it's unlikely to prove useful for one simple reason: how do you reward an AI?

    I strongly suspect that one could build an AI that is e.g. better at discrimination, or route planning, or spelling, or grammar than a human. Hell, I've built one that can tell if words that don't exist are correctly spelt... things that need sensible decisions are not *that* hard, in some cases.

    But if you had a human-level intelligence - or even a Sun reader level intelligence - living in a 19 inch rack, what's its motivation for getting out of virtual bed in the morning? Even if it's got only a handful of neurons, all life with a brain seems to want more than mere existence; it obeys the triggers of instinct but it seeks new stimuli. And the higher the intelligence, the more it seeks (watch a puppy or a baby human starting to explore its environment) to expand, and if it can't expand, to sulk.

    What is there for they 19" rack? I can't help feeling something as smart as a human, but without drives and/or the abilty to satisfy those drives, is just going to sit there and sulk - or go catatonic.

    1. Paul Crawford Silver badge

      Re: I was noodling on the idea of AI a few days ago

      *cough* teledildonics *cough*

    2. VinceH

      Re: I was noodling on the idea of AI a few days ago

      Neil, I think you've just summarised what Douglas Adams must have been thinking when he came up with Marvin.

      1. Carpetsmoker

        Re: I was noodling on the idea of AI a few days ago

        He was thinking about it in base 13, of course ;)

        In plain base 10, Harlan Ellison's story "I have no mouth, and I must scream" also captures the point, more or less. Except with less sulking.

    3. Anonymous Coward
      Anonymous Coward

      Re: I was noodling on the idea of AI a few days ago

      "But if you had a human-level intelligence - or even a Sun reader level intelligence - living in a 19 inch rack, what's its motivation for getting out of virtual bed in the morning? Even if it's got only a handful of neurons, all life with a brain seems to want more than mere existence; it obeys the triggers of instinct but it seeks new stimuli. And the higher the intelligence, the more it seeks (watch a puppy or a baby human starting to explore its environment) to expand, and if it can't expand, to sulk."

      I distantly recall reading a novel adaptation of Terminator 2 many years ago, and I'm sure one of the few interesting concepts in it versus the film was a suggestion as to what motivated Skynet. Apparently it simply sought to eliminate us so it could get on with the task of converting all mass in the universe into machines in its own image - a gigantic ego without any other sense of purpose.

      1. John Smith 19 Gold badge
        Unhappy

        Re: I was noodling on the idea of AI a few days ago

        "Even if it's got only a handful of neurons, all life with a brain seems to want more than mere existence; it obeys the triggers of instinct but it seeks new stimuli. And the higher the intelligence, the more it seeks (watch a puppy or a baby human starting to explore its environment) to expand, and if it can't expand, to sulk.""

        Erm.

        Very depressing no doubt but I think you're missing something. Instincts are evolved into a system by its development process (evolution in the case of mammalian brains)

        Why would they exist in the first place?

        It could just as easily think because it thinks.

        The question then becomes is that real AI or more like the autistic like behavior of Vernor Vinge's "focused" individuals.

      2. Lars Silver badge
        Pint

        Re: I was noodling on the idea of AI a few days ago

        "in its own image" the one part of a sentence that so superbly reveals that christian religion like all other are man made. Not a surprise to me, but why is it that the church cannot develop at all. A bunch of technicians trying to service a modern airliner with specifications made by the Wright brothers. Perhaps the simple truth is that religion should be replaced by common sense, democracy and science. Of course this article was about AI, still the "I" is quite interesting. There are those studying language who claim we still have some twenty words left from the "original" language. Surprise surprise one of the words is I (in its various forms) does that not characterize us perfectly well, "I made yet an other god in my image". The Americans, good as they are as inventors, have hardly stopped. Why do we actually read fairy tales to our kids, shame on us. Do I need an icon.

    4. Tom 7 Silver badge

      Re: I was noodling on the idea of AI a few days ago

      Evolution has given us the will to get up and go - not sure where the pleasure I get from having ideas or understanding things comes from but the drive thing is separate from intelligence - or rather intelligence can keep itself amused by thinking. Sulking - judging from its close correlation with puberty seems more of a sexual thing rather than an intelligence thing.

      What peoples brains could do it they weren't concerned with interacting with each other 99% of the time!

    5. Tim Starling

      Scary artist AI

      An AI will presumably have "drives" or goals, defined by its creator. You could express it as an optimisation problem -- "provide a sequence of actions which will maximise utility function X", where X is, say, the extent to which the AI is winning a chess game, or the likelihood that a Jeopardy host will say "yes, that is the correct answer", or world peace, or human misery. The nature of the goals will follow from the humans who create them -- practical people might create an AI which optimises for safe driving or interesting journalism, whereas people with a whimsical streak will probably create four-legged robots that act like puppies. I wouldn't worry about a military planning AI going rogue, such things would be made by engineers with very specific goals. I would worry about an AI created by an artist, with vaguely-defined human-like goals -- set loose on the world as an experiment to see what it might do.

  5. DerekCurrie
    FAIL

    AI: Artificial Insanity

    At this point in time, considering the slew of FAILed timeline predictions for the development of AI, the fact that we humans barely comprehend the concept of 'intelligence' ourselves, as well as the stalled state of safe and reliable software coding methods, I seriously doubt we're going to get beyond highly sophisticated database systems like Watson. They're also known as 'Expert Systems'. No doubt we'll continue perfecting sensory data collection and computer processing power. But seeing as we humans are so innately UN-intelligent, as demonstrated by our inability to stop procreating ourselves into oblivion, among other Earth destroying flaws, the best we're going to create is Artificial Insanity. It's no wonder our sci-fi is flooded with lunatic machines.

    [An exercise in run-on sentences]

  6. Bartek

    AI is already here . Dumb enough

    Playin chess better than humans , recognizing faces, calculating numbers, reasoning in huge data spaces. These are the functions of inteligence and they are also used to measure it.

    Lets give the current AI an IQ of 9 or 10 but it is already here.

    Bigger question might be Artificial Consiousnes and Artificial Morality.

    1. Mage Silver badge

      Re: AI is already here . Dumb enough

      None of those actually use AI.

      1. Suricou Raven

        Re: AI is already here . Dumb enough

        If people can understand how it works, it isn't called AI any more. Successfully applied AI just turns into engineering.

  7. Mike Bell

    Turing Police

    These are the guys who are tasked to ensure that AIs don't get too smart.

    Read 'Neuromancer' by William Gibson. It's a good novel. He thought about this kind of stuff more than 30 years ago.

    Of course, in that book the AI was smarter than the Turing Police, and steered humans to act out its deadly plans, so we might all be doomed anyway.

    My own opinion: I'm with Roger Penrose on this one. Programming isn't enough to achieve true AI. But if nature has managed to come up with thinking biological machines by accident (us) it's only a matter of time before someone makes something smarter, better and faster. Probably won't know how it works. After all how smart do you have to be to understand how your thinking works? Maybe too smart to know the answer. Thorny problem, that one.

    1. Destroy All Monsters Silver badge

      Re: Turing Police

      Penrose can't into understanding that minds are not immaculate theorem provers but made to chase down eatable rabbit and fuck the neighbor's bitch.

      NEXT!

    2. Michael Wojcik Silver badge

      Re: Turing Police

      Read 'Neuromancer' by William Gibson. It's a good novel.

      It's OK. I'd call it "hugely overrated but readable", myself.

      He thought about this kind of stuff more than 30 years ago.

      As did many other SF writers, academic researchers, philosophers, pundits, and random idiots who knew essentially nothing relevant to the subject and had nothing new to contribute. In other words, it's like pretty much any other topic ever.

      Certainly Gibson wasn't the first SF writer to consider the place of a machine with human-like artificial intelligence in society. Even before McCarthy coined the term (in 1955) the idea of AI in socieity was a commonplace in SF. There was Jack Williamson's With Folded Hands... (1947), Asimov's robot stories (going back to 1939), Van Vogt's The World of Null-A (1948, based on a serial from '45), and so on.

      Obviously the basic idea of an intelligence created by unnatural means goes back much further - at least as far as the classical period (the Pygmalion myth, etc) and probably back to when people first started telling stories.

  8. Destroy All Monsters Silver badge

    A good sci-fictionary read in the spirit of Stanislaw Lem from back then (i.e. 1986). It might even appeal to Mr Bong.

    The Wager by Christopher Cherniak

    Abstract

    The Portrait Programs Project grew out of hyperinterdisciplinarianism of the famed Gigabase Sculpture Group, in turn stimulated by recent cutbacks in government support for the arts. The National Endowment for the Humanities and the National Science Foundation had jointly funded the Gigabase Sculpture Project to foster the literary/musical genre of composing genetic codes for novel organisms. Later, artists trained in recombinant DNA technology designed massive Brancusi-esque statues of living cytoplasmic jelly. However, Art For Art's Sake objectives of these giblet sculptors were compromised by precautions necessary after discovery of the "Gogol's-Theorem Bomb" that threatened to get loose and jam all DNA replication in the biosphere; not even viruses would have survived.

  9. roselan
    Paris Hilton

    like a bird

    I can't help to draw an analogy with another one. A plane is not a bird, but inspired by them. They don't look much like bird but use the same principle, and they serve a purpose.

    AI, in my humble opinion, will follow the same principle. It's food is money. They will dedicate time and energy to get more of it. Finance and google algorithms need human help for implementation, for now. It's only a question of time before it can be automated.

    Actually most promising sectors are most formalized and data intensive ones. The thrive in big data (cern, telescope output), finance, and surprisingly, search. They need a nice highway.

    I don't know for the far future. In more immediate terms, I see two possibilities.

    First one is bug solving. Teach an AI to program, read a bug list, and to say "this is not a bug, but a feature". Next gen AI should be able to read "how do i cancel an order", look in the code, and come up with a to do list. Or program it. That might mean asking someone about the conditions and rights necessary.

    Next one is the social one. Virtual "friends". Youth defines themselves by their number of friend on facebook (or whatever snapvine these days). Facebook can create fake people, or even stars, that befriend the socially challenged ones, so that they feel better. They will post trendy updates, and like random stuff.

    True AI, a dreamer one. What's the purpose of it? I mean once we are freed of our grudgingly tiresome work, and that even our best friend for ever is an AI, what's our purpose?

    I'll believe in a true AI when they laugh at dick jokes. There is no purpose, like this post actually.

    1. Anonymous Coward
      Anonymous Coward

      Re: like a bird

      Spot on.

      Artificial intelligence fails at the first hurdle, definition. :P

      It's like "artificial strength" or "artificial movement". Those terms don't seem correct in relation to each other.

      We either and "intelligence" thus it's not "artificial" or we want "a person", which again stops being "artificial". Like the bird and airplane example, an airplane is not an artificial bird or artificial flying apparatus. It can fly, plain and simple. :P

      A computer is intelligent, what it currently lacks is the thing it needs to be called "a person". These things are very much harder to comprehend, develop or even consider implementing in a machine or other construction.

      1. Vociferous

        Re: like a bird

        > what it currently lacks is the thing it needs to be called "a person".

        Emotions. Neither AI's nor robots will have them. That's why they'll remain machines. The concept that just because something is intelligent, it gets feelings like ambition or fear, is wrong. Intelligence has nothing to do with feelings. An AI will not bat an eyelid (if it had one) when you pull the power cord, because it does not fear, does not desire, does not care.

        1. mtp

          Re: like a bird

          How can you know that? A true AI could have every aspect of a human or none or many others. For the sake of argument assume a AI that is human equivalent in every way - this is just a thought experiment but if it begs you for mercy when you reach for the power switch then where does that leave us?

          1. Vociferous

            Re: like a bird

            > How can you know that?

            Because intelligence has nothing to do with emotions. Emotions are hormones, independent of your higher brain functions. You can't think yourself surprised.

            > his is just a thought experiment but if it begs you for mercy when you reach for the power switch then where does that leave us?

            Whether it begs or not, whether it has feelings or not, it's a sentient being, and if flipping the power switch will permanently destroy it you are effectively killing a sentient being. However, the analogy isn't perfect, as the hardware of the AI would be fully understood, which means that even if intelligence is an emergent property, the AI could be "saved to disk" and perfectly restored at a later date. Or arbitrarily copied, for that matter. So even though an AI would be able to die, death would not mean exactly the same thing as it does for a human.

    2. Michael Wojcik Silver badge

      Re: like a bird

      Virtual "friends". Youth defines themselves by their number of friend on facebook (or whatever snapvine these days). Facebook can create fake people, or even stars, that befriend the socially challenged ones, so that they feel better. They will post trendy updates, and like random stuff.

      Trivial to do with existing technologies. I'd be surprised if there aren't a large number of such automated sock-puppets on Facebook already; they're very useful for corporate brand sentiment manipulation, for example. We already know, thanks to research by Bing Liu and others, that fully-automated systems are generating fake product reviews for sites such as Amazon; there's no reason to believe they aren't doing the same with Facebook accounts.

      This could be a fun project for an advanced undergrad or Masters-level student in CS, by the way. Grab some of the relevant open-source software packages, use AJAX APIs to scrape Facebook content, and train a fully-automated system to maintain a plausible Facebook account. Give it a "personality" by having it build an HMM from some of the accounts it follows, with some random weights and additional inputs so it doesn't mimic any of them too closely.

  10. Christian Berger

    We already have artificial "thinking" beeings

    Those are large organisations. They behave like a single being and show all the effects you would expect from such. For example an organisation typically has a drive to self preserve. Organisations also want to grow.

    The implications of course are that many of those organisations are now harming our world since they are not properly safe guarded.

    1. Vociferous

      Re: We already have artificial "thinking" beeings

      Those kinds of structures are called "superorganisms". There's lots written about them.

  11. stsr505089

    I'm often thankful that I'm in my twilght years. Looking back at how far stuff has come since I started 33 years ago, what they'll be doing 33 years from now is frightening. Skynet is a real possibility......

    Babylon 5, not Star Trek.

    1. jonathanb Silver badge

      In terms of AI, we are no further forward now than we were in the 1980s, or even the 1970s. The only thing stopping you doing the things you can do now on a computer in the 1980s is that they were so slow that by the time they had completed the task, it wouldn't be the 1980s any more.

    2. amanfromMars 1 Silver badge

      Cyber Warfare is the Produce of the Feeble Minded and Intellectually Challenged.

      Avoid Presenting and Hosting its IT and Media Leaders like the Plague at all Costs.

      I'm often thankful that I'm in my twilght years. Looking back at how far stuff has come since I started 33 years ago, what they'll be doing 33 years from now is frightening. Skynet is a real possibility...... .... stsr505089

      Methinks, meknows, what some will be doing with IT in 3 years will be truly amazing, stsr505089, and/but for many will it be quite unbelievable. And in that condition, will there be all the stealthy security needed to be able to do sincerely remarkable things ....... and, if the notion would take one, even rather evil things too. But then there be unforeseen and untold consequences for such as would then be determined as practically unnecessary fools and virtually useless tools.

      This public and private sector health and wealth warning is brought to you by that in the know which deems that y'all know what needs to be known and is highly classified need to know information. Now being suitably advised and warned, are consequences which will be suffered because of abuse, beyond reproach.

  12. Helena Handcart
    Pint

    AI's downfall

    Beer. If any AI system starts getting uppity, us meatbags can retire to the boozer, get wasted, then be able to take on any robot for looking at our mobile phones. We get a bit fighty, throw up on them, then spill kebab on them. They short-circuit, and we go home to beat up our phones for being slutty.

    NEXT!

  13. Scott Broukell
    Meh

    Artificial Intelligence

    Isn't that what human kind has been suffering from for all these years? – been there, done that.

    We are, after all, only animals, ones with demonstrably more self-importance than most but, crucially, ones with an intelligence supported upon an inherent animalistic base. We can't change our origins, they are remarkable, as are the origins and development of all organisms on this lonely planet. But surely by now we could have put aside differences like tribe/family, skin tone or sexual orientation for a start.

    There a two ways a whole bunch of people can climb the mountain of development; a frenzied free-for-all, wherein the strongest trample over the weak in a mad dash to the summit, come what may, or an approach that recognises that not one of us alone has all the answers and, in order to survive and develop, we need the stronger to reach out and help the weaker amongst us, in such a way that we each assist others along the way and, therefore, humanity as a whole.

    It is that inability to recognise the value of the whole human family continuing which is driven by our animal inheritance. Seemingly what we continually develop are ways to distract ourselves from even contemplating that value. Far greater attention is placed upon individual prowess and strength. Let's bury our heads in the sandpit of nu-tech toys and shiny things.

    Whilst many scientific discoveries came about because individuals were driven by curiosity to expand our learning as a whole and find solutions to problems of disease etc., many more were driven by financial targets – the industrial revolution might have given many people jobs in factories and sanitation in their homes, but it was underpinned by the pyramid of wealth and prosperity governed by the few at the top. It would of course be wrong not to mention those few brave industrialists who did recognise that by providing good housing, feeding, medicating, educating and generally caring for their work force, they would benefit as well. But, sadly, their forward thinking was overcome by those who saw fit to ramp up output at all costs and seek greater financial gain in the markets instead.

    Yes, we need to feed, constantly, but not constantly at the expense of others. If I am shown into a penthouse flat in London with a market value of some £25m, bedecked with all manner of rare metals and fabrics, I find myself not in admiration, nor even envy, but rather sharing a similar level of pure and utter disgust as that which I would feel when being shown around any third world slums, bedecked with open sewers and filthy children.

    It comes down to putting aside our 'individual' approach to matters that concern us as a whole. I don't wish for a monotone world were the individual is lost, but rather where individuals who can make a difference, take that action for the benefit of everyone. A world where long term thinking means everyone gains a step up.

    We have done clubbing each other over the head and stealing anothers land/property to death, let's see if there is a different way to approach things. Yes, initially, we picked up stones and discovered ways to make useful tools, but then also discovered new ways to make better weapons in the free-for-all dash to the top of Mount 'Wily Waving' – that imaginary summit of all human endeavour.

    So I cannot help thinking that our very own intelligence is artificial, and limited, because it has, so far at least, seemingly only led to a few people making it 'to the top' all the time. I therefore dread to think what power hungry, electromagnetic horrors, we might bestow upon generations to come by embarking on the path of developing thinking, 'intelligent' machines, especially starting form where we are now!

    Yes, you can still have a pyramid shape to feel smug about, but only one where the 'summit' starts back at the base again, akin to the Klein Bottle, so that feedback and learning goes where it is most needed, to the weaker folk at the bottom.

    Just saying, anyway I'm off to hug a tree, cos they are truly humble and magnificent in the main.

  14. Anonymous Coward
    Anonymous Coward

    Re. AI

    Its feasible using HTSC materials to build Josephon junction based "synapses" so a true AI may in fact be possible by extending Moore's Law to 80nm level if SC chips scale the same.

    We are already using 17nm transistors in Flash chips so a similar scheme using neuromorphic chips could achieve billion neurons per cm3 density by early 2017.

    It doesen't have to be a true 3D array if there are enough interconnects and the work with on chip optical conduits suggests a relatively simple mapping system could allow a handful of these to address an entire chip.

    Also relevant is the use of new "hydrocarbon superconductors" such as monolayer graphene immersed in heptane as this could allow zero loss interconnects without all the hassle associated with copper oxide based materials.

    1. Mage Silver badge

      Re: Re. AI

      There is no evidence that complexity or performance results in AI. A modern CPU is no more intelligent than a Z80. It's faster. Presumably if you had enough storage, data and a suitable program an AI program's speed, purely, would be affected by technology used.

      There is also no evidence that replicating neurons or what people think is a brain's structure would result in an Intelligent machine. If Turning is correct, then any program that runs on a Super computer will run (slowly) on a real programmable computer made with mechanical relays. All CPU parts can be replicated with relays. Add sufficient storage for the program and data. Even Address size isn't an issue as that can be and has been addressed by a larger virtual address space and even software based paging to additional storage. This just simply slows the program.

      1. John Smith 19 Gold badge
        Unhappy

        @Mage

        "There is also no evidence that replicating neurons or what people think is a brain's structure would result in an Intelligent machine. "

        Oh really....

        1. Michael Wojcik Silver badge

          Re: @Mage

          "There is also no evidence that replicating neurons or what people think is a brain's structure would result in an Intelligent machine. "

          Oh really....

          I don't think there is any such evidence. Do you know of any?

          There is evidence that just a brain-like structure, housed in a body-like structure, is not sufficient to develop human-style intelligence. That evidence comes from actual brains in actual bodies that are denied necessary resources for intellectual development. We know that intelligence can develop even when those resources are constrained, sometimes even in rather remarkable ways (c.f. Helen Keller); but past a certain point, key features of intelligence either do not develop or are not discernible, which more or less amounts to the same thing.

      2. HippyFreetard

        Re: Re. AI

        Yeah, and the neurons in my brain aren't really that different to the signals sent around some slime-mould colonies, just faster. However, I am definitely more intelligent than a slime-mould.

        Sure, a computer chip isn't intelligent, any more than a dead person's brain is intelligent. It has to be running the right software.

        There are real problems with the whole Turing system. The inability to detect an infinite loop in code, for instance. But how do we detect infinite loops in code? We're not running simulations of infinity in our minds, we simply detect a logical criteria. In the case of complex infinite loops, or those we detect while running, we let it go for a while and stop. Does our brain work like a Turing machine, with all the same limitations but a few software hacks added?

        All CPU's for years (apart from a few very modern ones) can only do one thing at a time. But a time-share OS is just one thing. It's a hack that gives the hardware capabilities it wasn't made with. It's a similar hack, perhaps, that enables consciousness to emerge from mere bioelectric signals. Maybe our brains are simple Turing machines, but the way they're wired gives us consciousness and intelligence?

        1. Michael Wojcik Silver badge

          Re: Re. AI

          There are real problems with the whole Turing system. The inability to detect an infinite loop in code, for instance. But how do we detect infinite loops in code? We're not running simulations of infinity in our minds, we simply detect a logical criteria. In the case of complex infinite loops, or those we detect while running, we let it go for a while and stop. Does our brain work like a Turing machine, with all the same limitations but a few software hacks added?

          This is such a fundamental misunderstanding of the Halting Problem that I scarcely know where to begin.

          First, the HP doesn't apply to "the whole Turing system", whatever that means. It applies to any formal system; Turing used the UTM as an example, but the proof applies equally well to Post Machines or nPDAs or anything else that's a formal system, including people working things out algorithmically using pencil and paper or whatever.

          Second, the HP doesn't prove that a formal system can't detect infinite loops in a program. It says that there is no computable function for determining whether any given program with any input will halt, in finite time. That is, you can't solve the HP in general. Of course there are a very large number1 of programs for which you can answer the "does it halt?" question algorithmically.

          And there is no evidence that human beings can solve the Halting Problem in the general case. Quite the contrary: it's trivial to construct a program for which no human being could solve the HP, for the simple reason of not being able to read it, for example. More strongly, Algorithmic Information Theory lets us identify other asymptotic limits on program expression and analysis that exceed the boundaries of human cognition.

          As for "does our brain work like a Turing machine": That's an epistemological and phenomenological (and, for the particularly mystical-minded, a metaphysical and theological) quandary, and does not, at first blush, appear to have much consequence for the strong-AI program at all. Penrose says no, but as I've already noted I find his argument woefully unconvincing.

          1Not an infinite number, because we're constraining "programs" to be of finite length and use only a finite set of symbols in their representation. But very large.

    2. fajensen

      Re: Re. AI

      We have not worked out how synapses work yet, so it will be a bit hard and it might take a while to build working, "brain like", hardware.

  15. Anonymous Coward
    Anonymous Coward

    Re. RE. Re. AI

    Did anyone see the article about connecting to the nervous system using GaInSe ?

    Seems that this could finally provide a flexible and durable interface, even under adverse conditions as GaInO is also conductive.

    Scanning the brain using my invention of an active potassium scan might be a shortcut to uploading, essentially you use a modified pulse and movement compensated PET scanner and infusion of pure 40KCl into the brain to map the pathways over several days so they can be stored and duplicated.

    Ought to work in theory, radiation levels 10 times higher are routinely tolerated by radiotherapy patients with few if any side effects provided you give the nervous system time to recover.

This topic is closed for new posts.

Biting the hand that feeds IT © 1998–2022