back to article Machine-learning models trained on pre-COVID data are now completely out of whack, says Gartner

Machine learning models built for doing business prior to the COVID-19 pandemic will no longer be valid as economies emerge from lockdowns, presenting companies with new challenges in machine learning and enterprise data management, according to Gartner. The research group has reported that "the extreme disruption in the …

  1. This post has been deleted by a moderator

  2. Anonymous Coward
    Anonymous Coward

    Puy du Fou is open

    UK, imagine a post covid world, similar to the pre-covid world, only with a few more masks and a bit more handwashing. All those stories of the 'post-covid' world will seem like a joke.

    Soon. Be patient.

    And when you're out of it, "Puy Du Fou" just emailed me, and they are open (with a few extra safeguards)!

    https://www.puydufou.com/en

    Do the quarantine, control the virus, and afterwards, well treat yourself to "Puy du fou". Trust me, I was right about the masks, I am right about this too.

    1. Doctor Syntax Silver badge

      Re: Puy du Fou is open

      No idea who they are and CBA to look them up?

      A/Cs are always right about everything. And always wrong.

      1. Anonymous Coward
        Anonymous Coward

        Re: Puy du Fou is open

        Brits have no idea what it is. It's like the best known secret in France. That's why I'm letting elReger's into the secret. But you have to clear Corona Virus first in the UK and be free to travel again.

        Patience. It's worth it.

        I don't think they're opening La Cinéscénie this year, but then you have to book that a year in advance anyway, its usually sold out.

        https://www.puydufou.com/fr/la-cinescenie

    2. Anonymous Coward
      Anonymous Coward

      Re: All those stories of the 'post-covid' world will seem like a joke.

      Ha Ha! Gran's dead!

  3. Ken 16 Silver badge
    Thumb Up

    Better invest

    In some strategy consulting then

  4. TeeCee Gold badge
    Black Helicopters

    Hmm.

    I sometimes wonder just how much kickback Gartner get from the Big Consultancies.

    After all, they're the ones who're going to get insanely rich[1] when world + dog asks them to fix the stuff Gartner says is screwed up.

    [1] Ok then, insanely richer.

  5. TheRealRoland

    But...

    Isnt this what a C-suite at a large company needs to be prepared for? That due to <reasons> markets fluctuate and how to react to that? Looks like boards of directors accepting that it's enough for someone to point fingers at datamodels specifically based on data that creates this market, to say 'i'm unable to carry out my work' ?

    Perhaps saying it diffierently: Shouldnt have models already accounted for market fluctuations due to issues like pandemics, natural disasters, etc.? If that's the case, someone is mailing it in.

    Where do i sign up?

    1. JDX Gold badge

      Re: But...

      ML learns from historical data to make predictions. If there isn't historical data for a certain scenario, it can't learn from it.

      This is the difference between human-designed algorithms, and data-driven machine learning (AFAIK).

      Since we haven't had anything like this in the modern, computerised era, humans are pretty much lost making predictions too. We'll have to wing it.

      1. Martin Gregorie Silver badge

        Re: But...

        If your fancy, data-driven predictive algorithm gets it spectacularly wrong just because the unexpected happened (pandemic, bank crash, military coup, whatever) then, ITS NOT A BLOODY AI, just a pattern matcher with insufficient predictive ability to deal with real world events. If the people who sold it to you claimed it was an AI and/or failed to specify its limitations and failure scenarios, then sue their arses off.

        1. Glen 1 Silver badge

          Re: But...

          " just a pattern matcher "

          Yes. You sound surprised.

        2. Shadowmanx2012
          Thumb Up

          Re: But...

          Thank you!

          At last, someone points out the "obvious" flaw in current A.I!

          1. John Brown (no body) Silver badge

            Re: But...

            "At last, someone points out the "obvious" flaw in current A.I!"

            Agreed, but I would point out that it's not the first time these flaws in so-called Artificial Intelligence or Machine learning have been referred to on these hallowed pages. I think pretty much all readers here would agree there is no such thing as "artificial intelligence" (apart from that poster who claims he i9nvnted it, of course!)

        3. James 139

          Re: But...

          If suffiently unqualified people call it an AI often enough, it becomes the "mass market" term for it.

          Like "information superhighway" to describe what, is now, very slow internets.

  6. DS999

    If (covid19_flaring && in_lockdown)

    then recommend(toilet_paper);

  7. Doctor Syntax Silver badge

    Why do I read this as Gartner saying "The dog ate my homework."?

    1. HildyJ Silver badge
      FAIL

      More than that

      Not only is Gartner "saying the dog are my homework" they are saying to all their subscribers "here is your Gartner certified excuse."

    2. veti Silver badge

      That was my thought too.

      I'm currently living in a post-covid-19 country, and I'm here to tell you it looks very much like the pre-covid-19 economy. There's a lot fewer tourists and a bit more unemployment, but hardly a quantum shift in customer demand for anything.

      1. ST Silver badge
        Boffin

        > I'm currently living in a post-covid-19 country [ ... ]

        There is no post-covid-19 country. There is no vaccine and there is no known cure for COVID-19.

        1. Brewster's Angle Grinder Silver badge

          I initially downvoted that but then changed my mind. Even the country which is Covid-19 "free" still has border restrictions affecting the economy.

    3. maffski

      Try reading it as 'people who sell models insist you need to buy models'

  8. YetAnotherJoeBlow

    AI?

    So... their models are not AI - they are just matching patterns. If it was AI, the systems would adapt themselves to work with the new data stream. In the mid-eighties scientists could do pattern matching like this, but the hardware was not up to snuff to scale with PDP-11's with 1MB RAM or even the mighty VAX.

    While some of the math has changed (more optimized), the end result is the same. It is all hardware driven. Now the problem will shift back to software since we no longer have exponential growth in CPU speed - proving yet again that our industry needs to do better than just software bloat and feature creep.

    I am very fortunate that I got my degree when I did. What I learned then is still very much applicable to my work today, and that trait is sound judgement.

    1. imanidiot Silver badge

      Re: AI?

      Given that "AI" is just a buzzword given to Machine Learning algo's and pattern matching systems, yes, they're just matching patterns. There ARE no proper AIs. They're all still dumb as bricks most of the time.

      1. John Brown (no body) Silver badge

        Re: AI?

        "They're all still dumb as bricks most of the time."

        That's a very good analogy. You can make all sorts of very useful things out of bricks. Bricks are very useful and come in different styles, shapes and materials so they can be used in different situations, On the other hand, you'd not really use bricks to build a car or an aircraft or even a ship, let alone a computer.

        And so, likewise, algorithms are dumb, most of the time. They can do certain things very well. But they are not truly "learning" or "intelligent".

  9. osakajin Bronze badge

    Maybe try to get the data from post spanish flu era? Now that is a real pandemic not like this excuse for jolly good thrashing of ordinary folk and wealth transfer that is currently happening.

    1. spacecadet66

      The reasons why data from the interwar period might not be the best basis for a contemporary prediction model are left as an exercise for the reader.

      1. osakajin Bronze badge

        twas a joke.

  10. gerryg

    Those who don't learn from history...

    In the 1950s RAND created "scenario planning" as a way of helping the US military develop "futures". This crept into MBA syllabuses as a tool for business strategy and corporate planning. If your AI cannot do scenario planning it's unclear to me why you are using it.

    1. amanfromMars 1 Silver badge

      Re: Those who don't learn from history...

      In the 1950s RAND created "scenario planning" as a way of helping the US military develop "futures". This crept into MBA syllabuses as a tool for business strategy and corporate planning. ...... gerryg

      Seventy years on, gerryg, and things have moved on significantly but are also still very much the same as they were whenever the almighty power of media was first realised and harnessed. Today though it is not a lonely and exclusive operation.

      Here is a Current ITeRation? .......

      amanfromMars [2006231238] ..... being more forthright on https://www.independent.co.uk/news/business/news/bank-of-england-insolvency-coronavirus-andrew-bailey-a9579216.html

      Eh? Where are you coming from? Are you talking Mars language? .... Arminius

      Howdy, Arminius,

      Have you not yet realised how everything works out there in the wild and do you think it is truly only available to a choice few rather than more than just many ?

      Words create, command and control and destroy worlds .... for are they not what everyone reacts to, imagining what are just personal opinions and/or collectively shared views in/on/for media programs to spin as prime important leading advice for the futures ahead.

      Question more as you have done, and you will be both amazed and terrified in equal measure with what can be easily found hidden away out of regular sight in dark secret places and corrupted spaces.

  11. Anonymous Coward
    Anonymous Coward

    most organisations are just starting to operationalise their machine learning efforts.

    I'm sorry, did you just use the word "operationalise" in cold blood?

    Jesus Fucking Christ!

  12. Mark192 Bronze badge

    Hard to plan for the future when you have to make decisions based on assumptions with no valid historical data to back it up.

    Businesses that have capable staff that know their market (and have the maths skills to do their own calculations) will have a big advantage over those relying on "AI" models.

  13. Warm Braw Silver badge

    Product recommendation engines or next-best-offer

    In my experience, they've never worked.

    That's historic learning that can reliably continue to be applied.

  14. Anonymous Coward
    Anonymous Coward

    Ah.....intelligence again!

    George Osborne.

    Dominic Cummings.

    David Cameron.

    David Davis.

    Dominic Raab.

    Liam Fox.

    *

    Remind me again what we know about HUMAN INTELLIGENCE........

    *

    ......before you blather on about the defects of the ARTIFICIAL kind!!!!!

  15. Spanners Silver badge
    Boffin

    Pattern Recognition

    Criticising supposed AI as "just pattern recognition" ignores a major fact.

    Most human intelligence is pattern recognition. For example, my younger colleagues consider me very good at mental arithmetic. I just know my times tables up to 16 (older Scottish) and so many other patterns that I don't even know they're there.

    All the people asking about the "post Covid world" are asking for new patterns. When early humans went to a new place, they were in more danger until they identified the patterns of the area - what certain birds flying in circles meant (vultures?), what particular shadows and noises indicated (predators that might eat them) and so on.

    Pattern recognition is intelligence.

    1. Anonymous Coward
      Anonymous Coward

      Pattern Recognition is Intelligence.......

      @Spanners

      ....until one stumbles across problems of "recognition".....

      *

      ....like mistaking a correlation for a cause....

      *

      ....and umpteen other mistakes of logic.....

      *

      ....all of which are EASILY embedded in so called "ARTIFICIAL INTELLIGENCE"....

  16. sketharaman

    They say past results are no guarantee of future performance. Then they use AI/ML systems to improve future performance. But these systems can't work unless they're trained on past data. #GoFigure

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

Biting the hand that feeds IT © 1998–2020