michael_nielsen

I was told in the past week that writing books isn't a good way to get info to the avg #nlproc person because the avg CS student doesn't read books---only web pages, I guess. (& videos.) I'd love for that claim to be proven wrong. If it's not wrong, how can we change this?

As much as I would like to consider books superior, the only times I have found them useful in #nlproc are when they are part of a larger course. In most other ways (accessibility, interactivity, shareability) I have preferred to learn from online tutorials or videos.

Also, I guess it depends on whether you consider something like this a book or a web page :) http://neuralnetworksanddeeplearning.com  It hits most of my sweet spots for how I prefer to learn new stuff.

What kinds of content are you learning from online tutorials or videos though? Does it include the kind of information from other disciplines that is required to do linguistically informed #nlproc, not to mention to engage with ethical issues?

That's where I find books superior, since (good) books provide a larger context to specific problem areas. When I search for information, I typically am interested in knowing how to solve a very specific issue, e.g. I need to relearn how backpropagation works for the nth time.

That rarely includes a larger interdisciplinary context, and whenever it does it feels rather shallow or sometimes even wrong. That's why I in parallell try to keep up with the current discourse happening within the community on Twitter or in conferences.

(Not saying this is the best way, it's just how my life currently works.)

And in such a context, would you make time to read a book?