back to article El Reg was invited to the House of Lords to burst the AI-pocalypse bubble

To Westminster, England, where the House of Lords is conducting a wide-ranging inquiry into artificial intelligence. I'm writing a book on AI – and why people believe in it – and on the basis of some notes, was invited to give oral evidence on a panel of three alongside the BBC's Rory Cellan-Jones and The Financial Times' …

  1. Bronek Kozicki

    nice

    Kudos for this "Find people who find flaws in an approach, and support researchers who have interesting new approaches that aren't popular or fashionable. Encourage intellectual diversity."

    I agree, that's what we need instead of following one-and-the-same path as everyone else.

    1. Commswonk

      Re: nice

      Kudos indeed, but it's hardly a new concept. Go and spend a moment or two searching for Edward de Bono's Six Thinking Hats and you'll find that you need the "Black Hat". (But don't throw the others away; they are also useful!)

      1. TRT Silver badge

        Re: I agree, that's what we need instead of following one-and-the-same path as everyone else.

        Yeah, it works until everyone starts doing that.

    2. Jonathan Richards 1
      Thumb Up

      Re: nice

      > Encourage intellectual diversity

      Maybe it would help if university academics didn't all have to compete *quite* so hard for research funds from the same committee that has been bamgoogled by the latest secular saint.

      Good work. Keep it up.

    3. This post has been deleted by its author

    4. This post has been deleted by its author

    5. Rebel Science

      Re: nice

      I agree. Unfortunately, the Deep Learning people don't want to hear anything that contradicts their paradigm.

  2. Anonymous Coward
    Anonymous Coward

    "Then Rory said something interesting...."

    Holy shit, that's a first.

    1. Anonymous Coward
      Anonymous Coward

      Re: "Then Rory said something interesting...."

      Is that why they only invite him on the Today program once a month?

      Mind you, Today is pretty boring in it's own right, year in year out - Trump, Brexit and terrorism ad naseum. Yawwwwwwn.

  3. Anonymous Coward
    Anonymous Coward

    "if this wasn't like a householder objecting to these newfangled aeroplanes flying over"

    Flawed analogy. If instead flying over at an enough high altitude - and only that - they would fly continuously over my property with a battery of sensors to gather data about it and everything/everyone inside, and process them to profile me as much as they can, well, I would object a lot.

    Maybe these are the same people who would object a lot if paparazzi stationed nearby their properties, or fly over it taking shoots of their private affairs. Time to send some drones around them, I guess, after all the photons you reflect outside your property and in the air are free for everybody to capture, right?

    1. Andrew Orlowski (Written by Reg staff)

      Re: "if this wasn't like a householder objecting to these newfangled aeroplanes flying over"

      That's excellent. Everyone should use this if confronted with the analogy again.

  4. Mage Silver badge
    Happy

    Brilliant

    Even harder than the amazing things a five year old can do, which we have no idea how, is changing the perception of politicians. Harder to get sensible decisions than true AI?

    I can't see any advance in AI since 1980s and the slow HW claim was a 1960s or 1970s fallacy. If the SW existed and HW was too slow, you'd get real AI, but slowly. Real AI doesn't exist and might never exist. It might be a class of problem we can't even analyse, much less engineer an implementation.

    1. Anonymous Coward
      Anonymous Coward

      Re: Brilliant

      I think computer games are a perfect example. Give any type of AI a set of rules to "play" in, and it may become the best.

      In the real world? We either don't know the rules to feed the AI, or must trust it can adapt given that the "rules" change (our needs and results required), though the laws of physics do not.

      A valid example: https://www.rockpapershotgun.com/2017/10/11/paperclips-browser-clicker-game/ (AI to the extreme)

      1. This post has been deleted by its author

        1. Anonymous Coward
          Terminator

          Re: Brilliant

          But unlike a person, a computer is still putting a car on a track...

          If we assume we can copy "intelligence" then fine, we can do anything. But it is an assumption. At this point, we can do algorithms, and we can do math (busy beavers ;) ).

          So so far, all evidence points to us being able to make a more perfect algorithm, an more perfect train track taking us to a destination. But how do we make a train that can choose any detonation it wants?

          1. This post has been deleted by its author

            1. AdamWill

              I was gonna say...

              ...it should be nominated for 'Freudian slip of the year'...

    2. This post has been deleted by its author

  5. SkippyBing
    Terminator

    Sarah O'Connor

    With a name like that how could she possibly be in favour of AI?

    1. defiler

      Re: Sarah O'Connor

      I see I was beaten to it. I was worried things had gone off the rails around here!

    2. really_adf

      Re: Sarah O'Connor

      Obviously, they looked up the wrong name in the phone book.

  6. amanfromMars 1 Silver badge

    Uniting Fronts for Casting Webs.

    Nice one, Andrew O

    But invited to burst the AI-pocalypse bubble, Andrew, or decided to introduce Lords and Knights of the Realm to ITs Sublime Presence providing and provisioning Future Augmented Virtual Reality ...... with ProgramAble Futures and Derivatives easily made available Live for Live Operational Virtual Environments ........ for and with Brave New Orderly World Orders Exercising HyperRadioProActive Quantum Communications Control for Universal Command.

    When the Lords have been told by people with great authority that everything is on the right track, how can you introduce this idea? I'm not sure I succeeded.

    If you need more help with that sticking point, Andrew, provide them this comments thread.

    You know, there are moments in life when you want to stick a fork through your head, then you realise you are on live TV. This was one of them. There wasn't a fork handy.

    Priceless. We are AI Mused. :-)

    Does El Reg host and field Red Team Players? Would it like to?

    And that all says pretty much everything as I know IT is so far ........ We Think, We Are One and Almightily Creative Together.

    Cc Global Command Head Quarters for Cheltenham Harriers re Virtual Space Missions with Quantum Communications Ports of Call with Heavenly Postings?.!

    1. amanfromMars 1 Silver badge

      Re: Uniting Fronts for Casting Webs.

      And be this something of an opposition or simply novel foreign competition? ....... https://9to5mac.com/2017/10/11/apple-original-content-headquarters/

      Have you had any engagement with those planning things there, Andrew?

      1. m0rt

        Re: Uniting Fronts for Casting Webs.

        Kinell, Aman!

        I thought you were dead/shutdown/disconnected. Glad you aren't.

        But I agree with your sentiment. AO - I really enjoyed your musings on your take on this entire debacle. I wish you had stated your mind, instead of self censoring due to the, suposed, gravitas of the particular company, but I fear I have to unblock ad blocking again for the week.

        Well done.

    2. Alistair

      Re: Uniting Fronts for Casting Webs.

      AMFM:

      I see you've read 2312 lately. Wonderful book.

  7. John Smith 19 Gold badge
    Unhappy

    Minsky's "Weather analogy" pretty much nails the whoel "Deep learning" business.

    Other classics are "regression equations" derived from data sets.

    Algorithm chomps on X MB of data and comes up with N term equation. Plug in X value, get Y value within Z %

    But WTF does that mean?

    What are the real parameters (that could be measured) to give the dependent value?

    And does this equation hold outside the test range? No idea. We don't have a data set for that.

    Before you think this sounds nuts a common model for turbulence in CFD sims is the "2 factor" model," where you twiddle 2 numeric values.

    They have no connection to the actual physical reality of the problem being modeled.

    1. Mephistro

      Re: Minsky's "Weather analogy" pretty much nails the whoel "Deep learning" business.

      That's the bad news. The good news is that by creating and analysing these simulations and their failures, scientists get hints of what other factors are affecting the weather, so, at least in theory, all this "parameter wrangling" can be slowly eliminated while, at the same time, scientists gain a better understanding of the 'real world climate' and simulations of a better quality.

      1. m0rt

        Re: Minsky's "Weather analogy" pretty much nails the whoel "Deep learning" business.

        The way I see it is this:

        You have one entry point, and three exit points in a maze. There are millions of purmutations allowing one through. You observe a cockroach going into the single entry point over thousands and thousands of iterations and predict which exit they come out of.

        You may be able to predict to a reasonable degree, using 'deep learning', which exit they come out of based on observation of the entry. But you wil never know why.

  8. Phil Endecott

    Neuroscience

    My guess is that gaining a better understanding of how brains work would be one of the better ways to spend AI money.

    1. TRT Silver badge
      Boffin

      Re: Neuroscience

      Yeah, yeah, yeah! Kerchiing!

    2. breakfast Silver badge
      Boffin

      Re: Neuroscience

      It goes deeper than that - the problems of AI are not just about how brains work, they are about the underlying philosophical questions regarding the nature of knowledge and consciousness. Many of the greatest minds of the last three thousand years have explored this and still not got to any definitive answers, so assuming that we can ram a bunch of information into a big database then run some statistical rules across it and come to any useful conclusions is perhaps a trifle optimistic.

      1. TRT Silver badge

        Re: Neuroscience

        Mmmmm..... *drool*

        Trifle.

  9. Pete 2 Silver badge

    The AI engine

    Consider the electric motor. It is ubiquitous and invisible at the same time. It is embedded in maybe half the things that have a power supply (leaving out light bulbs). But while it performs the same action - turning electricity into movement - it's function in each appliance varies.

    Although factories (back in the day when they were called manufactories used to have one honkin' great source of motive power: a water mill, steam engine, whatever. Now they rarely do: they have many dedicated sources of physical movement where that application is needed. So it is with computers - mainframe verses embedded. And so it will be with AI.

    The AI of the future will be invisible. It won't be a giant "brain" watching over us. Deciding the fate of humanity or the planet. it will be lots of little devices making decisions, adapting them, optimising. Maybe working together as we network computers now. Each one doing a little bit where it is needed. More worker bees than Terminators.

    The main issue we will have is in developing the tools to program these little AIs. The compilers and editors if you like. What degree of abstraction will the devices work with? How will we define their goals (rather than their operation - that is the AIs job). How will we keep them on the "straight and narrow" - not wandering off in some undetermined direction like Sirius Cybernetics lifts. And how will will get them to recognise when they are being hacked?

    1. This post has been deleted by its author

    2. Long John Brass
      Terminator

      Re: The AI engine

      @Pete 2

      Yes; I'm not sure why more people don't get this...

      Intelligence at the edge of a big dumb network; However our lords and masters only ever see and want a centralised command and control. IE a big smart AI sitting in the middle of the network.

      I guess that has more to do with the personality types of the people you are talking to that any kind of over-arching reality.

  10. defiler
    Terminator

    Sarah O'Connor and AI?

    Come with me if you want to live.

  11. tom_slee

    Book!

    You say: "I'm writing a book on AI". Great stuff: look forward to it.

  12. Richard Tobin

    Dustbin of history

    Trotsky, not Lenin.

  13. Alistair
    Windows

    Artificial Intelligence in the real world.

    Many many many moons ago I read "A door into Summer". And quite a few other books that featured barely independent, partially independent, moderately independent, and completely independent mechanical devices. Most of these were "Robots" that made humans lives simpler, easier, more efficient and less work.

    When we here and now speak of "Artificial Intelligence" we (especially us TechnoGeeks) have an expectation that AI will be in the class of completely independent devices. I'm thinking that our path to that device will be long and difficult and will start with the lowly roomba. Sadly however, a roomba will still fall down the stairs if you don't fence them off, and can still manage to get itself jammed behind the couch. And *that* is where AI as our society sees it is stuck. It is the roomba stuck behind the couch.

    There are tasks that are done that can be taught to purpose built devices and then repeated in perpetuity by that device. What we have great difficulty doing is teaching that device how to cope with something that we have *NOT* taught it directly - and *THIS* is the pivot point. Until the roomba can figure out, solely from observation, and understanding of its own limitations, and comprehension of the environment around it, how to move the couch and get out from behind it, our ability to create an AI will stagnate at feeding in enormous amounts of iterative data and *averaging* that (either literally or using algorithms) to mimic "intelligence". No matter *how* large the pool of input data, there will be future events that *do not* fit either the iterative knowledge or the algorithm.

    Andrew: is there a public record anywhere of this committee? Love to see it in full.

  14. trashsilo
    Thumb Up

    House

    House of Lords has been pretty resilient to employment changes over the years, maybe the right people to ask about future employment change and keeping a job.

    The UK certainly punching above their weight in terms of parliament member numbers alright.

    Second highest in the world at 1430. China first on 2987 and Italy in third spot manages a paltry 950 or so.

    Source : Wikipedia List of legislatures by numbers of members.

    Quality and interesting article Andrew, found myself hanging on your observations in the audio.

    You asked 'How to start?' - wow, you came bang out all hype sceptic, knocking some straight over.

    Spot on with the liability, accountability and privacy issues.

    Universal income not a premature argument but overdue for me.

    Maybe there has already been an initial "Release the algorithms!" moment in finance and marketing,

    with not many really seeing past the non-transparent transparency.

    The data will be exploited for marketing purposes. The supplementary aeroplane flying over your house analogy and wanting a fee in royalty is not tight. The AI data agents be they smart phone, meter, cookies, cctv, wearables are flying fast 24x7 and some may not want a fee they may need a off button.

    A look over the Rio Tinto, BHP self driving mining trucks suggests they manage a very impressive park unaided.

    Grasp the AI opportunity UK, perhaps the biggest risk is bumbling along class lines and being irresponsible with the data.

  15. amanfromMars 1 Silver badge

    The Sad Rad Truth? Some are more Equal than Others who are Retarded?

    Have you concluded, Andrew, most present lords and ladies are not fit for Future Leading Greater IntelAIgent Games purpose?

    Or do you imagine them able to be Beta Enabled and made Sensitive and Receptive to HyperRadioProActive IT for Supply and Support in Live Operational Virtual Environment Fields?

    Why wait for such stragglers and strugglers when to move on ahead unilaterally without them, preparing and presenting the Novel Ways to be Followed, will provide them freely and remotely with everything they need to feed and seed.

    Surely you cannot deny or disagree that all that is needed is AIMagic Script Supply?

  16. Anonymous Coward
    Anonymous Coward

    The biggest threat isn't the AI itself

    but our ability to think being eroded.

    "computer says no" will become normal and accepted. there are already examples of people following dumb devices like sat-navs onto railways, off cliffs, into fields etc.

    AI is not creative, does not handle new situations well but we will all trust it. Suddenly all data about behaviour will be based on machine learning too, thus ensuring it will tell you what you want to hear, rather than what is going on.

    AI in itself can be useful, but the danger is that it is assumed to be capable of being rational, and having (increasingly un-)common sense..

    1. This post has been deleted by its author

  17. Anonymous Coward
    Anonymous Coward

    AI is the future, just like the Paperless Office circa 1992...

    1. allthecoolshortnamesweretaken

      Paperless Office circa 1992

      Oh, I heard that at the 1985 CeBIT (the last year that CeBIT was "just" the computing part of the Hanover Fair before becoming a gig in it's own right).

      Oh, and AI was "just around the corner" then as well.

  18. Robert D Bank

    Intelligence is not consciousness

    so without consciousness AI is only an I/O supervisor, making decisions in isolation based on it's inputs. It can iterate many possible outputs and learn which might be optimal within certain parameters, but is not aware whether those might not be optimal for those using it for whatever purpose. If a self driving car for example decides to drive off a cliff to save driving over 30 cyclists that cannot be avoided, is that optimal or not? From a conscious drivers point of view a whole slew of decisions would flood their mind, taking into account a lot of 'unconnected' things that an AI wouldn't. Should I sacrifice myself for the greater good? Should I steer into the rock wall on the other side so I might have a chance of surviving? Do I actually care about the cyclists that much?

    Just a thought.

  19. Doctor Syntax Silver badge

    "He thought the problem today was that software was clever and hardware was slow."

    The problem - or at least part of the problem - for AI is that the hardware RI runs on is massively parallel. Whatever counts as massively parallel in silicon is nothing compared to what the brain runs on.

  20. Anonymous Coward
    Anonymous Coward

    The AI naysayers

    I am convinced that these people won't acknowledge that AI is a threat until it becomes sentient, which may well be a long way off.

    The disruptive effect of widespread AI adoption will have had a massive impact long before that occurs.

  21. Anonymous Coward
    Anonymous Coward

    The sooner politicians are replaced by AI, the better.

    The UK would benefit enormously from it.

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