back to article Neural networks: Today, classifying flowers... tomorrow, Skynet maybe

If there's a poster child for machine learning, it's neural networks. We gave a practical introduction to the topic here, but this time I'll take a different approach and explain the background to how neural networks, er, work. To demonstrate this, we'll show how a neural net can be used to classify different species of iris …

  1. Tom 7 Silver badge

    "Once we have a set of weights that seem to work"

    That's the fun bit. You harden that up and then another outlier arrives...

  2. Nifty Silver badge

    This article still doesn't explain why the 'Like That Garden' app on IOS works in such a vague way. The old gaffer down at the allotment can identify plants better!

    1. Anonymous Coward
      Anonymous Coward

      Because classifying between only two specific types of iris is WAY easier than classifying all the various plants you might find in a garden?

    2. Anonymous Coward
      Anonymous Coward

      "This article still doesn't explain why the 'Like That Garden' app on IOS works in such a vague way"

      Maybe it's meant for Lady Gardens?!

  3. Doctor Syntax Silver badge

    I suppose the next we hear they'll have invented cluster analysis just like we had 50 years ago.

  4. Anonymous Coward
    Anonymous Coward

    An interesting article. I learned something about an area I want to play with some time in the not too distant future.

  5. handleoclast

    Back propagation, back propagation, back propagation

    I said it three times, and thus I have summoned...

    Except this is probably the one time he won't spring out of the woodwork to say what a crock of shit he thinks back propagation is.

    1. Mage Silver badge

      Re: Back propagation, back propagation, back propagation

      It's nothing to do with neural as in real brains.

      Dishonestly named and isn't part of AI. It's just a way of building a database using human curated examples.

  6. MonkeyCee

    Standard AI issue

    Obligatory xkcd.

    And for good measure:

    Look, I love me a neural network as much as the next lazy AI creator. I love them so much that I build them on GPUs, so you can hide quite how much buggering around you did to get to a sensible answering system.

    But like all AI problems, if there are clear goal states and easily classifiable middle states, it's a "very easy" to solve problems. Hence why we'll keep finding AIs that can play at an advanced adult level in games (which are artificially constrained) but can't manage child like tasks.

    An AI usually needs to have already seen something in order to classify it. So a child/AI knows what a fire engine is, because it's seen lots of pictures of them, seen one going past, had it explained to them what it is. If you then say "at the airport, fire engines are yellow" you can be pretty sure the kid will figure out without more explanation, but the AI classification may have decided that something bein red meant fire engine ore than any other input.

  7. Anonymal coward

    Plus ca change, plus c'est la....

    In the late 70's at Brunel, we had Igor Aleksander and his neural networks. I think the only thing that's changed since then is the horsepower available to do this stuff; there certainly seems to be no qualitative advances...

  8. PhilipN Silver badge

    great, great, grandson

    I think you mean great-great-grandson. Unless you wanted to say great, great great-great-grandson, which sounds sycophantic.

    And by the way : imagine you are a cyborg programmed with machine learning and ask yourself "why?" such as "why do I want to get out of bed in the morning?" or even "Who gives a toss about these stupid flowers?"

    Crikey with a semblance of humanity most days I (with certain physiological needs) do not want to get out of bed in the morning.

    Skynet? Pfffffttt!!

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