back to article Human intelligence may not be enough: US military turns to machine learning algos to predict food shortages

The US Department of Defense is building machine learning tools to help predict critical food and medicine shortages as America grapples with the coronavirus pandemic. Supply chains are buckling under the strain of COVID-19 as meat processing plants, factories, and warehouses wind down their operations or shut up shop as …

  1. IGotOut Silver badge

    It's not a problem.

    A rich white man in ultimate power is forcing poor people back into meat processing plants, regardless of risk to their health.

  2. Anonymous Coward
    Anonymous Coward

    I wonder if this A.I. will be clever enough to tell them that when people don't have the money to buy food it doesn't matter if it's produced? There's a storm coming and removing the lock down will only make it worse.

    1. John Brown (no body) Silver badge

      Or, for that matter, be able to predict a sudden panic buying spree. People may seem calm and relaxed, but once reports come in of empty store shelves in one location, the mass panic will quickly spread. We've already seen it happen once in recent weeks.

  3. Nifty

    Didn't they know that even the smartest algorithms and AI predictions will get Trumped?

  4. HildyJ Silver badge
    FAIL

    Data?

    Beyond the problems of food production, supply, and affordability, the big question is where are they going to get the data? The reason FEMA had to physically go to grocery stores and count supplies is that grocery stores aren't going to turn over their inventories and orders to the Feds any more than Apple is going to tell them the breakdown of iPhone models sold or their chip and screen orders.

    1. Smooth Newt Silver badge
      Megaphone

      Re: Data?

      Beyond the problems of food production, supply, and affordability, the big question is where are they going to get the data?

      Don't worry about those tedious little practicalities. It's Artificial Intelligence, and so will Save Us All. That's all you need to know.

  5. holmegm Bronze badge

    "representing over a million people in retail, meatpacking, pharmaceutical, and manufacturing industries, estimated that at least 72 of its members had died from COVID-19 so far."

    I'm not a math wiz, but maybe that isn't quite as scary as you seem to think that it is? How long is "so far", for one thing?

    1. Anonymous Coward
      Anonymous Coward

      In this state of 5m people there have been <100 deaths all in nursing homes and with a median age of >85

      So 70 deaths of healthy working age manual workers is worrying. Especially since not all the million people he represents will all have been working, so the denominator is less.

    2. Rich 11 Silver badge

      Given that the US COVID-19 death rate "so far" is approximately 190 per million head of population, that's not bad going, at least on the surface. Of course the wider figure won't be purely for those of working age, but it would still help if they were able to provide a valid comparison.

      1. Smooth Newt Silver badge
        Meh

        Given that the US COVID-19 death rate "so far" is approximately 190 per million head of population, that's not bad going, at least on the surface. Of course the wider figure won't be purely for those of working age, but it would still help if they were able to provide a valid comparison.

        A 190 per million mortality rate depends on continuing to prevent the majority of the population from getting it. No one knows what proportion of people have had it, maybe 5% as a wild guess? If so, and it spread across 100% of the population, and they all only got it exactly once, then the mortality rate would be 3.8 deaths per thousand.

        1. Yet Another Anonymous coward Silver badge

          I'm not super sure you can rely on US numbers in the same way as for countries with a national health service (and a functional government)

          Are they 100% testing homeless deaths or even people who die outside hospitals?

          Is the coroner office / cause of death reporting in small towns as rigorous as in NY?

          Once this breaks our of big cities into rural areas where fewer people have health care, the health care infrastructure isn't as great and people are less able/willing to stay at home - it could get a lot worse

          1. Smooth Newt Silver badge
            Terminator

            I'm not super sure you can rely on US numbers in the same way as for countries with a national health service (and a functional government)

            I don't think you can rely on any country's numbers. They are too easy to massage.

            The only tolerably accurate measure is to compare the total number of deaths from all causes now with the same numbers from last year, because it's very hard to fiddle that.

            1. Rich 11 Silver badge

              because it's very hard to fiddle that

              But expect the politicians to try.

          2. holmegm Bronze badge

            All the incentives are actually to inflate the death numbers. Funds from higher to lower levels of government, insurance coverage, private charitable efforts, all sorts of things incentivize reporting *higher* death numbers.

            Died from a heart attack while also testing positive for C19? A "C19 death". For example.

  6. Ken Moorhouse Silver badge

    AI's Conclusion...

    Along the X axis: number of deaths.

    Along the Y axis: amount of food manufactured.

    Objective: How to get a straight line graph with a positive gradient?

    Solution: S_yl_nt Gr__n

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