back to article Our intuitive AI outperforms (most) puny humans, claims MIT

Traditionally computers are great at crunching numbers, but lousy at understanding what they mean. But a team of researchers at the Massachusetts Institute of Technology thinks it's cracking that problem. A new paper, to be presented at next week's IEEE International Conference on Data Science and Advanced Analytics, details …

  1. Ole Juul

    mind reading still required

    The MIT team suggests that the computer system could be used to establish a baseline of results to check against a human response.

    Sounds good. Now we just need a way to make meaningful data out of all the stuff that happens in a human's head in response to something.

    1. Fraggle850

      Re: mind reading still required

      >Now we just need a way to make meaningful data out of all the stuff that happens in a human's head in response to something.

      That's not necessary. Any stuff that happens in a human's head is ultimately manifested through their subsequent actions and this is what is being analysed from the subsequent data trails that they leave. That is the whole point of such technologies.

      1. Anonymous Coward
        Anonymous Coward

        Re: mind reading still required

        That is the whole point of such technologies.

        Well let's see them put it to some use. I propose the Ledswinger test, in which a machine qualifies for AI status only when it comes up with a coherent and realistic plan for resolving the complete mess that is the Middle East at lowest cost and minimum inconvenience (without resorting to WMD or "final solutions").

        1. Naselus

          Re: mind reading still required

          "I propose the Ledswinger test, in which a machine qualifies for AI status only when it comes up with a coherent and realistic plan for resolving the complete mess that is the Middle East at lowest cost and minimum inconvenience"

          It'd come back with '2 state solution' after about eight seconds.

          Now waiting for Davedavedave to declare me antisemitic for that statement.

        2. LucreLout

          Re: mind reading still required

          @Ledswinger

          coherent and realistic plan for resolving the complete mess that is the Middle East at lowest cost and minimum inconvenience (without resorting to WMD or "final solutions").

          Big wall? (I discounted putting ecstacy in the water supply every day and burning copious quantities of weed on the evening, due to the likely religious objections from both sides.)

          Realistically there is no way to resolve it - too many generations have killed each other over the same bit of turf. You're never going to see eye to eye and live peacefully with someone that blew up your children. [1]

          To achieve lasting peace, you'd need a few decades with nearly no killing such that relations could normalise. No amount of talking to or discourse between these people [2] will achieve that - we have no words left unspoken over the preceeding generations and they have not achieved peace. Dialog will not deliver.

          I've reluctantly come to the conclusion that you may as well spend your time trying to design a perpetual motion machine, as sorting out that conflict.

          Peace in the ME, if it is to come, must come from them - it is not in the gift of the west to bestow. And that desire for peace will only become stronger than the mutual hatred once the killing and the loss become unbearable, and for that they would have to be allowed escalate to truly sickening levels [3].

          So all we're left with is extreme solutions, or continuation. Minimising the impact on the rest of us more or less comes down to a very big very high wall - nobody in, nobody out, no news reports in or out either. Pop a drone over the top every 100 years or so and see if peace has happened yet.

          Down voters, for inevitably they will come, please enlighten me as to your better solution, for I am genuinely interested. It isn't a problem AI can solve, because it isn't a problem perpetuated by lack of intelligence applied to it.

          [1] I've been very careful not to mention race or religion or sides here. For the sake of argument lets just assume it applies equally to everyone.

          [2] Still applies to all involved.

          [3] Still with the applies to all sides gig. No dsicrimination here.

          1. Anonymous Coward
            Anonymous Coward

            Re: mind reading still required

            "Peace in the ME, if it is to come, must come from them - it is not in the gift of the west to bestow."

            And given that the most extreme elements there are driven by deities, they would rather go MAD than sing Kumbaya. Tragically, we'd sooner see World War III than genuine peace here since the differences are too fundamental. It's hard to argue with someone who feels he/she belongs to the chosen race and are destined to survive Armageddon...

            1. Crazy Operations Guy

              Re: "most extreme elements there are driven by deities"

              Not so much, it more has to do with people claiming to represent those deities for the sake of claiming money and power. Do you think that Netanyahu and his cabinet believe in God? Or Abu-Bakr Al-Baghdadi actually cares about the Koran?

              The extremist leaders only use religion as a means to an end rather than a guiding principal. A head of state can blame every failing of their own policies by saying that the people aren't pious enough, that bridges collapse because the devil did it, or that the people you want them to kill are sub-human abominations unto your god. Rather than you being incompetent, insufficient funding for infrastructure, and those other people are perfectly reasonable humans that you happen to hate.

              1. Anonymous Coward
                Anonymous Coward

                Re: "most extreme elements there are driven by deities"

                I'm not blaming the heads of state, who to some degree have to be rational in order to be a leader of men. I'm talking about the extremists behind those leaders, some of whom really are both mad and willing to go MAD.

                1. Anonymous Coward
                  Anonymous Coward

                  Re: "most extreme elements there are driven by deities"

                  Mandatory intermarriage: you can only marry someone from the other side. Pretty soon, there is only one side.

          2. DocJames

            @LucreLout Re: mind reading still required

            we have no words left unspoken over the preceeding generations

            I, respectfully, declare that bollocks. There's been very little talking between the sides compared to any other complex dispute (and this one is, as you point out, extremely complex).

            OTOH, I'm not sure how we could get them to start talking more. Perhaps some expectations from those that support each side wouldn't go amiss. (I'm looking at you, USA and Iran). And the rest of the Arab states not being so bloody minded in 1) hating Israel almost as much as 2) hating the Palestinians and leaving them in the dirt.

            And apart from that, I completely agree with you, especially your point that "[p]eace... must come from them".

            1. LucreLout

              Re: @LucreLout mind reading still required

              @DocJames

              I, respectfully, declare that bollocks. There's been very little talking between the sides compared to any other complex dispute

              http://www.bicom.org.uk/resources/timelines/peace-plans/

              37 talks, plans, resoloutions, and solutions. Yes, they could always talk more, but realistically, after 48 years to talks, they've said what they have to say. Though my "big wall" solution doesn't preclude them talking amongst themselves while inside the wall.

              I doubt everyone who has lost their children to the conflict will accept a bunch of politicians from both sides declaring peace - and lets face it, any political solution will fail at the very next attack. Conflict may grant them some form of revenge, while peace brings them nothing they value now that they have no children, no future.

              Fraggle mentioned "miracle dust in a can", which seems about as likely to bring about lasting peace as more talks.

              1. DocJames

                Re: @LucreLout mind reading still required

                37 talks, plans, resoloutions[sic], and solutions... after 48 years

                This is why I say compared to any other complex dispute. They've not really been trying consistently; they've been waffling intermittently, when prodded. I agree that "a bunch of politicians declaring peace" ain't going to work - it requires real commitment from [almost] all on both sides. That's why serious talking is required. Those in Norn Ireland who have lost loved ones seem to have come to an acceptance that having arguments over how to organise the recycling is better than ongoing violence - a significantly less complex conflict, but still one over decades with many deaths on both sides. I don't know what value they attribute to peace compared to having lost children; as you point out this is a real problem (and not one solved by outsiders - including politicians - imposing "the answer"). It does seem to have been mainly solved however.

                And I agree research into miracle dust might also be cost effective in finding a ME solution. Sadly.

          3. Kamal Hashmi

            Re: mind reading still required

            Yep. True peace always comes bottom-up not top-down - but the top needs to be receptive.

            Of course, war can come bottom-up or top-town.

        3. Fraggle850

          @Ledswinger Re: mind reading still required

          > a coherent and realistic plan for resolving the complete mess that is the Middle East at lowest cost and minimum inconvenience (without resorting to WMD or "final solutions").

          It's called 'artificial intelligence' not 'complete and utter fucking miracle dust in a can'. It is becoming more effective than humans in some well defined cases where there is significant data but that is not the nature of this particular problem.

          Going off topic: I think that the solution requires various state actors to stop politicking and start having respect for the ordinary people who live in that region. Dropping the religious rhetoric would also help. It's only function seems to be to enforce division through tribalism.

        4. Michael Wojcik Silver badge

          Re: mind reading still required

          I propose the Ledswinger test, in which a machine qualifies for AI status only when it comes up with a coherent and realistic plan for resolving the complete mess that is the Middle East

          Why ask more of AI than we do of the sum of human intelligence? Well, why not, I suppose - the whole idea of such a test is largely pointless anyway.

          Of course your comment would be more apropos if the article had anything to do with AI.

      2. Ole Juul

        Re: mind reading still required

        @Fraggle850: That's not necessary. Any stuff that happens in a human's head is ultimately manifested through their subsequent actions and this is what is being analysed from the subsequent data trails that they leave.

        The stuff that happens in a human's head is the stuff of humans. To exchange what is "ultimately manifested through their subsequent actions" with the world of humans is an obfuscation and I fear that the developers have used an old trick here. I enjoy a magic show as much as the next guy, but after the show I go back to the real world where humans and physics are no longer the same.

        1. Fraggle850

          Re: mind reading still required

          To quote Arthur C Clarke: 'Any sufficiently advanced technology is indistinguishable from magic'

          If, as the researchers claim, the technology performs better than humans on specific tasks then there is no trickery and whatever they are using as inputs to the AI serves its purpose.

  2. Pascal Monett Silver badge

    "what they've done is going to become the standard quickly"

    As soon as it can be monetized, that is.

    And if Big Data is anything to go by, it will be by this time next year.

  3. James 51

    Sounds like something that would help with analysing the results of medical trails, as long as it had all of the data and what it had could be trusted :P

  4. Fraggle850

    Won't be long before they apply these techniques to business

    This sort of thing could make swathes of manglement obsolete at a stroke. There are going to be some early adopters who realise their mistake too late, followed by resistance as other managers see what is happening.

    AI appears to be becoming mainstream, machine learning is available in MS Office now. I guess that cloud and general computing power, coupled with big data, have finally got us to a place where the long held promise of AI and its related techniques can be realised.

    Be interesting to see how this pans out, these truly are interesting times for AI and machine learning.

    1. Stevie

      Re: Won't be long before they apply these techniques to business

      This sort of thing could make swathes of manglement obsolete

      Depends on one major factor: Whose engrams were used.

    2. Tom_

      Re: Won't be long before they apply these techniques to business

      "This sort of thing could make swathes of manglement obsolete at a stroke."

      So you're saying it's going to be of little to no consequence?

    3. Naselus

      Re: Won't be long before they apply these techniques to business

      "This sort of thing could make swathes of manglement obsolete at a stroke."

      Which is why it won't be applied to business. Automation is to replace the proles, not the suits.

      1. Fraggle850

        Re: Won't be long before they apply these techniques to business

        Ah yes, but the unfortunate thing for those in the lower and middle echelons is that those above them consider them to be proles too.

  5. Named coward

    In other news: "computers can crunch numbers faster than humans" (by all means a very interesting paper, especially the part where the autotuning worked better than the initial manual parameters)

  6. DropBear
    Facepalm

    Oh great - we finally have machines that can do Rorschach tests. Still dumb as a brick, mind you, but now a brick that can be psychoanalysed! Progress FTW...!

  7. amanfromMars 1 Silver badge

    Ye Olde PEBKAC Enigma and Abiding ZeroDay Vulnerability Exploit Portal for Virtual Machine Masters

    A machine may easily be able to highlight a problem and suggest a logical solution but that does not mean that humans will follow the advice whenever it upsets vested interests hell bent on retaining and sustaining a lucrative inequitable advantage that delivers excessive perceived reward/fiat riches/apparently anonymous and virtually remote, politically incorrect command and control/arrogant media maintained manipulation of ignorant masses.

    And Virtual Machine Masters are an expanding explosive phenomena/futures trading initiative/ab fab fabless opportunity and certifiable nightmare of an intangible problem for hands on resolution in recalcitrant and degenerate established vested interest circles/clubs/orders.

    Life with AI is a Great Head Game and that is what you have to deal with, for to ignore it and the power of exploring ubiquitous IT and Mass Media ReProgramming will see IT takeover and makeover everything in images of ITs own choosing, with no discernible input from supposedly in command and control human administrators/exclusive executive systems operators.

    cc GCHQ [your Ref TJ245] …..Subject and Object Lesson Nothing much hidden there to bamboozle and entertain Communications Technology Analysts.

    1. Brewster's Angle Grinder Silver badge

      Jealous, much?

    2. TitterYeNot

      Re: Ye Olde PEBKAC Enigma and Abiding ZeroDay Vulnerability Exploit...

      MIT's Data Science Machine spotting trends and patterns in large datasets much faster than humans is all well and good, but I'm not going to be really impressed until it manages to decipher the hidden meaning in amanfromMars1's posts...

      1. LucreLout

        Re: Ye Olde PEBKAC Enigma and Abiding ZeroDay Vulnerability Exploit...

        .. but I'm not going to be really impressed until it manages to decipher the hidden meaning in amanfromMars1's posts

        But surely their impeneterability is intrinsically linked to their quality? Each post not only means something different to each man or woman that reads it, it means something different to each man or woman each time they read it!

        1. Fraggle850

          Re: Ye Olde PEBKAC Enigma and Abiding ZeroDay Vulnerability Exploit...

          So amanfrommars is a management consultant then?

          1. Sir Runcible Spoon
            Thumb Up

            Re: Ye Olde PEBKAC Enigma and Abiding ZeroDay Vulnerability Exploit...

            "So amanfrommars is a management consultant then?"

            Quite aside from that being totally unbelievable and probably quite insulting, it did make me laugh :)

            Have an upvote.

            1. Fraggle850
              Pint

              Re: Ye Olde PEBKAC Enigma and Abiding ZeroDay Vulnerability Exploit...

              Upvotes all round, your response made me chortle too.

              I do wonder which party you are suggesting might take offence though?

  8. Whitter
    Devil

    Does this mean the "crazy Amazon recomendation game" is on again?

    The recent recs are so tedious: mainly stuff I just bought 5 mins ago!

  9. Ru'

    No doubt the year of AI will be running on all the Linux desktops...

    1. Pedigree-Pete

      AI on Linux...

      Ru' Perhaps the AI will run on the paperless Linux Desktops promised in the early 80s.

  10. Amorous Cowherder
    Boffin

    Aha!

    You see Data Science Machine....DSM...rearrange that and you get MSD....which are the initials of the wife of the part time electrician who work on the set of the movie "Ooooh Missus Ermintrude, There's A Sausage in My Mulligatawny!" which was being made on the same lot as 2001: A Space Odyssey!

  11. Anonymous Coward
    Anonymous Coward

    Or to put it another less dystopian way...

    ... a group of very smart researchers and programmers developed algorithms that can outperform other people in this particular sphere.

    The intelligence isn't in the machine, its in the people who programmed it. You could - in theory - get a human who knows nothing about data pattern recognition to follow the algorithm with pencil and paper and outperform other people as far as the end result goes. Though not I suspect in the time it takes!

    1. Destroy All Monsters Silver badge
      Thumb Down

      Re: Or to put it another less dystopian way...

      The intelligence isn't in the machine, its in the people who programmed it.

      That kind of vitalist bullshit is trotted out regularly as the last refuge of the denialists (I found that they also start to blabber about God or equally bizarre concepts in the next minute) but it's not true.

      - Your intelligence is not in your kid.

      - Organizational know-how is not in the CEO or the investor

      - The kinetic energy of a locomotive is not in any of its parts

      - Capacity for Big Data analysis is not in any of the programmers or even the team

      1. Fraggle850

        Re: Or to put it another less dystopian way...

        Well put, have an upvote. People read AI and instantly start thinking of it in terms of human intelligence. It's nowhere near that (yet - who knows where it could go in the future). This isn't about developing an artificial consciousness (whatever that might be) it's about utilising the current state of the art to do things more efficiently/effectively than humans can.

        There was an excellent Marcus du Sautoy documentary on BBC4 recently that looked at current uses for AI. The AI driven Occado logistics operation was a wonder to behold.

      2. Anonymous Coward
        Anonymous Coward

        Re: Or to put it another less dystopian way...

        "That kind of vitalist bullshit is trotted out regularly as the last refuge of the denialists (I found that they also start to blabber about God or equally bizarre concepts in the next minute) but it's not true."

        Whatever it is you're smoking I suggest you partake somewhat less of it.

        Its a machine. It doesn't have any intelligence, it just blindly follows a list of instructions. The instructions were written by one or more programmers. THEY are the ones with the intelligence.

        Educate yourself:

        https://en.wikipedia.org/wiki/Chinese_room

        1. Fraggle850

          Re: Or to put it another less dystopian way...

          I have (re) educated myself with regards to the Chinese Room and note that the wiki article states:

          > it is not an argument against the goals of AI research, because it does not limit the amount of intelligence a machine can display

          Now, can we quit with the philosophical masturbation and accept the article for what it is: using well defined AI techniques has been demonstrated to outperform humans in some cases.

          Yes, the people who write the code are very clever. No, no one is saying that machines have developed consciousness. But the fact of the matter is that AI, once coded, trained appropriately (assuming it's an AI that needs training) and set to task is essentially a black box. You don't know what it's doing to arrive at its decisions.

          AI is in the realm of the engineers nowadays. It will be used more and in different ways as we become more comfortable with it. There will be errors, overreach and unintended consequences but that is the nature of progress.

          It is reasonable to say, even within the confines of Searle, that AI systems have some intelligence (for some subset of what we believe to be human intelligence - even that is not fully defined).

          1. Sir Runcible Spoon
            Terminator

            Re: Or to put it another less dystopian way...

            "Its a machine. It doesn't have any intelligence, it just blindly follows a list of instructions."

            In that case there are surely large swathes of the population that should be re-classified.

          2. Anonymous Coward
            Anonymous Coward

            Re: Or to put it another less dystopian way...

            "You don't know what it's doing to arrive at its decisions."

            Sorry, thats simply not true. For current computer technology , if you know whats in memory, the code thats going to be run and the input you can predict *exactly* what the output will be. There is nothing random about a von neumann computer - even the pseudo random number generators are predictable. It might be hard but its perfectly possible.

            1. Fraggle850

              Re: Or to put it another less dystopian way...

              I did consider qualifying my statement to that effect but decided not to for reasons of brevity and clarity. My point is that the AI is determining how to solve the problem with which it is presented, the operator doesn't know the details of what it is doing, even if said operator built the AI. You could look under the hood and analyse what you see it doing and it is all deterministic. I certainly didn't claim that the process had any randomness!

            2. Someone_Somewhere

              Re: Or to put it another less dystopian way...

              "f you know whats in memory, the code thats going to be run and the input you can predict *exactly* what the output will be."

              Not if you use a gen-alg, you won't.

              Nor if you use LISP properly.*

              --

              *Nobody understands lambda calculus - which is why LISP was invented in the first place ;)

        2. Michael Wojcik Silver badge

          Re: Or to put it another less dystopian way...

          Educate yourself:

          https://en.wikipedia.org/wiki/Chinese_room

          Educate yourself. Searle stated, in print, that he believed machine intelligence was possible and would quite probably be developed eventually, because he believed the mind was a mechanical effect of its physical substrate (the body).

          The point of the CR thought experiment was to argue against one particular trend in strong-AI research at the time, what Searle referred to as "symbolic manipulation".

          Searle had no patience for the sort of mysticism you're promoting.

          1. Anonymous Coward
            Anonymous Coward

            Re: Or to put it another less dystopian way...

            "Searle had no patience for the sort of mysticism you're promoting."

            Me promoting mysticism?? Hello pot , meet kettle! I'm not someone who believes conciousness is going to arise "somehow" out of current computer architectures. You're all deluded wishful thinkers. Its certainly possible with some physical systems otherwise we wouldn't be here, but not through what we call a computer today. Go read up on Godels incompleteness theorum as to why. And if you still don't understand, well, just go back to your dreaming.

            1. Fraggle850

              Re: Or to put it another less dystopian way...

              I'd be interested in how you are applying Godel's theorems to AI.

              I personally have no strong opinion on whether machines will ever achieve consciousness but am prepared to accept the possibility that they might. Unless you believe in some animistic force then you must admit that we are physical systems. If we are just physical systems then our sense of self must arise from those physical systems. If we could replicate those systems exactly then it should develop consciousness. Maybe we can emulate it in software at some point in the future, maybe it will be on a vastly different architecture. Perhaps, in some distant future, our current efforts will be viewed in the same way we may regard the early chemists' phlogiston theory.

              That philosophical discussion has little to do with the matter at hand though. The article is about using techniques that have come from attempts to model intelligence. The techniques apparently perform some specific tasks better than humans. It is fair to say that the machines exhibit some aspect of intelligence, albeit somewhat limited.

            2. Michael Wojcik Silver badge

              Re: Or to put it another less dystopian way...

              Me promoting mysticism??

              Yes, that's exactly what you're doing. Watch: in a minute you're going to make a handwaving reference to the Incompleteness Theorem as if it somehow supports your argument.

              Hello pot , meet kettle! I'm not someone who believes conciousness is going to arise "somehow" out of current computer architectures.

              Who claimed it would? Not me. I didn't even claim machine consciousness was possible; I simply noted that Searle did, because you so grossly misrepresented his position.

              You're all deluded wishful thinkers

              Oh, and you don't think that bit of foot-stamping is anything more than an avowal of faith on your part?

              Go read up on Godels incompleteness theorum as to why

              I assume this is some pitiful, sophomoric attempt to mount an argument along the lines of Penrose's. You're no Penrose, and I don't find his version convincing in the first place. (Notably, he fails to demonstrate that the human mind can do anything with incompleteness that a deterministic mechanism cannot, because he violates a level of abstraction: between formal incompleteness and the perception of formal incompleteness.)

  12. Destroy All Monsters Silver badge
    Go

    In honor of William Gibson

    I should be called "Blue Ant" or "Bigend"

  13. Stevie

    Bah!

    Very good.

    Now let's see it do it after a week-long power cut and consequent oil products shortage.

    8oP

    1. Anonymous Coward
      Anonymous Coward

      Re: Bah!

      Sure, soon as you can demonstrate the human number-crunchers achieve the same feat without resorting to coffee. Or sleep.

    2. Fraggle850

      @Stevie Re: Bah!

      I'm assuming that you are an IT person? What computer can do anything with the power off? Your argument seems to be: "anything that requires electricity is useless because of its dependency" I fail to see how that is relevant.

      1. Stevie

        Re: @Stevie Bah!

        That is my point. Or was. I've moved on.

  14. Anonymous Coward
    Anonymous Coward

    Computers are good at working with numbers shocker

    News at 10 - Sky is blue, Water is Wet, Earth is round.

    1. Someone_Somewhere

      Re: News at 10 (was 'Computers are good at working with numbers shocker')

      "Sky is blue"

      Or is it?

      http://thesocietypages.org/socimages/2012/12/28/is-the-sky-blue/

      http://thesocietypages.org/socimages/2012/08/05/language-culture-and-color-a-visit-with-the-himba/

      I used to know a couple who managed to convince their son (until the age of seven) that the reason some older films and TV programs were in black and white was because colour wasn't invented until the '50s :D

  15. Anonymous Blowhard

    "We view the Data Science Machine as a natural complement to human intelligence,"

    Data Science Machine: [dreaming] Kill all humans... must kill all humans... kill all humans...

  16. Primus Secundus Tertius

    Object-oriented abstract noun

    I am curious to see how an abstract noun - understanding - might be modelled in an object-oriented programming system.

    1. Michael Wojcik Silver badge

      Re: Object-oriented abstract noun

      Well, you've poisoned the well by giving a (dubious and imprecise) grammatical concept - abstraction of nouns - undue ontological status. Claiming that abstraction is a property of understanding per se, rather than "understanding" the English word, is a very tenuous and fraught position.

      Second, you're making by that gloss an implicit phenomenological claim, which is that understanding as an activity or state of mind is in some sense "abstract" rather than "concrete". But a mechanical theory of mind would dispute that, and say that understanding qua the process of knowing and working with a concept (in some more-or-less "correct" fashion) is a concrete, material activity - signals in the brain or some such.

      But if we hold those two points in abeyance for a moment, and consider just the modeling question you proposed, the obvious answer is to use the same procedure we'd use for any other physical-world concept we want to model directly as an OO artifact. We'd consider what attributes an "understanding" has in our system, and what want to be able to request of it (methods or messages, depending on your preferred paradigm).

      The attributes would probably 1) identify the understanding (describe the concept it applies to), and 2) map its relations with other understandings and concepts in our system. (Hard-core epistemologists can argue over whether an understanding is purely relational, a la Saussure, or has some inherent referential meaning, a la Pierce.) For messages/methods, we need at least some sort of dispatch: apply this understanding to this situation and produce an interpretation.

      Implementation is left as an exercise for the reader. I don't recommend C++.

  17. Kubla Cant

    When you start thinking about machine solutions for real-world problems, it becomes clear that much of the cleverness lies in defining the problem domain.

    Take something slightly simpler than the conflicts in the Middle East: the Great Heathrow Runway Debate. If you're going to ask a machine to resolve that question, then you have to assign values to all the conflicting interests. Having done that, you've probably come to your own decision without the machine.

    1. Fraggle850

      @ Kubla Cant

      > much of the cleverness lies in defining the problem domain

      Correct

      > assign values to all the conflicting interests. Having done that, you've probably come to your own decision without the machine

      Partially correct. Whilst the solution may be arbitrarily binary, the variables aren't necessarily so clear cut. Assigning values is part of the solution but how individual data points are classified based on a number of different variables pertaining to each data point is where the AI comes in. The AI will determine it's own neural weightings and classify accordingly. If it is set up correctly then it may come up with a non-intuitive solution. Obviously I'm referring to classic neural networks here, I'm not fully up to speed on the latest developments in the field.

  18. Fraggle850

    AI is coming to a workplace near you

    For all the (occasionally confused!) debate here AI is a maturing technology: get with the programme or get the f**k out of Dodge.

    Here's a scary thing: the HR department WILL be using AI. They will use it to decide whether to hire you and they will use it to monitor your performance. It is quite likely that some are already doing so, I've certainly seen AI used to determine whether potential hires are likely to stick around long enough to make it worth while taking them on. That was a theoretical use case but the results looked fairly compelling.

    1. DocJames
      Happy

      Re: AI is coming to a workplace near you

      You haven't met my HR dept. I'd be happy if they learnt to use spellcheck.

      1. Fraggle850

        Re: AI is coming to a workplace near you

        I think I've worked alongside your HR department, that inability to use spellcheck sounds depressingly familiar... just don't go pulling them up about it, they'll send you on a course so that you can learn how to appreciate your colleagues' creative approach to language.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like