back to article That AI scanning your X-ray for signs of COVID-19 may just be looking at your age

Machines are like humans – they’re lazy. When given the chance to take the easy route to complete an easy task, they will. Academics at the University of Washington found that algorithms trained to diagnose COVID-19 from chest X-rays often look at secondary features, such as a patient’s age, rather than focusing on the images …

  1. amanfromMars 1 Silver badge

    Mouse traps for the control of vermin?

    With particular regard to MICE and DARPA interests and requirements, the solutions needed there may be held securely otherworldly ...... for there is too much of a similarity and singularity in the report above, "DARPA wants to fund AI research studying information warfare" ... filed by Katyanna Quach Mon 7 Jun 2021 // 04:06 UTC and the intel shared freely here within the El Reg posting ..... https://forums.theregister.com/forum/all/2021/06/04/biden_admin_issues_ransomware_guidance/#c_4269260.... to not be worth mentioning and/or issuing an advisory warning about should one be prone to consider it problematical to accept rather than an unusual opportunity to explore and exploit wholeheartedly.

  2. Anonymous Coward
    Anonymous Coward

    Those weapons sound familiar.

    Wasn't there a previous generation of weapon that was left in place to loiter and eventually kill a target. I believe they were known as Land Mines.

    1. Muscleguy Silver badge

      Re: Those weapons sound familiar.

      Except land mines are indiscriminate. They are only non lethal to the side which lays them if they remember where they are accurately. Also they sit there killing and maiming civilians for decades. Drones cannot, yet, stay up for decades rendering areas no go.

  3. Gene Cash Silver badge

    Problem with "AI" in general

    "but the association is unexpected and not transparent"

    That seems to be the problem with all of the AI out there. It gets "trained" but not on what we thought it was training on. Kind of like the WW-II dogs that were supposed to deliver bombs to enemy tanks. They attacked friendly tanks because they didn't train on the shape the tank but the smell of the diesel.

    1. amanfromMars 1 Silver badge

      Re: Problem with "AI" in general

      "but the association is unexpected and not transparent"

      That seems to be the problem with all of the AI out there. It gets "trained" but not on what we thought it was training on. Kind of like the WW-II dogs that were supposed to deliver bombs to enemy tanks. They attacked friendly tanks because they didn't train on the shape the tank but the smell of the diesel. .... Gene Cash

      That's where AI Scouts have got IT sussed, Gene Cash. They are immaculately prepared for the cloudy unexpected ..... living to love striving and thriving in IT as one does.

      Every moment a delight to exploit and expand upon ....... and share with all, inviting Engagement and Astute Agile ACTive Participation in JOiNT Future Leading Seeding Operations.

      Methinks that is not an AI Problem whenever an Enigmatic Human Dilemma for Satisfactory Resolution ........with/in an Attractive Solution/Heavenly Answer.

  4. Robert Carnegie Silver badge

    That film where they are about to missile the pub where brown terrorists may be, but there's a kiddy in the way. This is a comedy presumably? In real life it would be a five minute play? Including theme music.

  5. Muscleguy Silver badge

    Always look under the hood

    If your AI’s method are not transparent, you cannot look under the hood, then what you have may well be a shortcut lemon. In real world parlance, it might run but it might have a blown gasket, a shot muffler, a clutch that is just hanging on and a leak in the fuel tank.

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 © 1998–2021