5 ms·

I wonder if students that take this class are better at throwing their pile of data into Tensorflow.

minihat 5y ago
This cynical point of view is shared by a number of engineers I know. Another version of it is 'why is it worth learning the calculus of machine learning when that is mostly abstracted away by Tensorflow/PyTorch/JAX?' To a software engineer accustomed to operating on layers of abstraction far removed from the hardware, this may seem a reasonable point. Why is it worth learning that pesky math, anyway? I would argue that the machine learning engineers of today are more like electrical engineers than programmers, however. When something goes wrong, you don't have nice warning messages or error catching available to you. Like an electrical engineer with a voltmeter, one must begin probing inputs and outputs each step of the way. Good luck doing that if you do not understand how the components are supposed to work. YMMV by copying and tweaking others code, but I believe we are still far off from hands free 'autoML'. Just ask anyone who has sent a model to deployment whether AWS autoML was sufficient for them. And whether they needed someone who understands backprop at some point during the model training process.
rg111 5y ago
That works only for cats vs. dogs toy classifiers. It had never ever worked for anything practical. And no, Deep Learning is high science. It is not just throwing piles of data to PyTorch and run it on a 8-GPU cluster. I don't blame anyone having this view. Megacorps with billions of bucks are in a race to train the biggest language models, and that is the news every media focuses on.
- clircle 5y ago
  Deep learning is definitely not science
  - rg111 5y ago
    We still don't know how exactly things work inside. But we will, someday. We, once considered the existence of "unknown forces" in basic planetary motion. But we figured out later. There are many directions, one of them being the work being done by Bronstein et al. w/ Geometric Deep Learning.
    - clircle 5y ago
      planetary motion is a property of the natural world. Deep learning algorithms are not. Deep learning is not science.
      - fuy 5y ago
        In this sense computers and language, for example, are also not a property of natural world. So you can't have sciences that study and/or use computers or languages?

This cynical point of view is shared by a number of engineers I know. Another version of it is 'why is it worth learning the calculus of machine learning when that is mostly abstracted away by Tensorflow/PyTorch/JAX?' To a software engineer accustomed to operating on layers of abstraction far removed from the hardware, this may seem a reasonable point. Why is it worth learning that pesky math, anyway? I would argue that the machine learning engineers of today are more like electrical engineers than programmers, however. When something goes wrong, you don't have nice warning messages or error catching available to you. Like an electrical engineer with a voltmeter, one must begin probing inputs and outputs each step of the way. Good luck doing that if you do not understand how the components are supposed to work. YMMV by copying and tweaking others code, but I believe we are still far off from hands free 'autoML'. Just ask anyone who has sent a model to deployment whether AWS autoML was sufficient for them. And whether they needed someone who understands backprop at some point during the model training process.

rg111 5y ago

That works only for cats vs. dogs toy classifiers. It had never ever worked for anything practical. And no, Deep Learning is high science. It is not just throwing piles of data to PyTorch and run it on a 8-GPU cluster. I don't blame anyone having this view. Megacorps with billions of bucks are in a race to train the biggest language models, and that is the news every media focuses on.

clircle 5y ago
Deep learning is definitely not science
- rg111 5y ago
  We still don't know how exactly things work inside. But we will, someday. We, once considered the existence of "unknown forces" in basic planetary motion. But we figured out later. There are many directions, one of them being the work being done by Bronstein et al. w/ Geometric Deep Learning.
  - clircle 5y ago
    planetary motion is a property of the natural world. Deep learning algorithms are not. Deep learning is not science.
    - fuy 5y ago
      In this sense computers and language, for example, are also not a property of natural world. So you can't have sciences that study and/or use computers or languages?