Data Science in Context – Peter Norvig's New Book
datascienceincontext.com
datascienceincontext.com
Brilliant.
The single biggest issue I see working in this area today are two things:
Lack of historical context. People have no clue about the history of quantitative management (rocky, lots of ups and downs)
And sales. Terrible, terrible salesmanship from nerds.
Anyone involved in this field professionally needs to read this book.
Rapidly.
I am definitely going to give this a read. After decades of more or less using GOFAI, the last eight years has mostly been machine learning and deep learning. Lately I have been scratching an itch to combine old fashioned symbolic AI with more modern deep learning (hybrid systems). I started a new job on Monday where I think this may happen.
From the table of contents, the four examples look appropriate for looking at ML in the context of large real world problems.
This is also quite practical for large consultancy firms. Most of the chapters, I’ve had clients discuss with me (such as Responsible AI). Personally, I think it could have went away from the applications as it was too high level.
Data scientists are not software engineers.
They analyze data, they don't produce applications.
Not everything in the world is written for the benefit of software engineers.
But in the book, computation is one of the three braids comprising data science. And they include software engineering in computation.
See book sections 1.2, 1.2.4, 2.1.3 etc
—
What I am trying to untie is, should data engineering be distinguished from data science, or not?
This is better than yet another machine learning algorithm book.
Disappointingly, to me, this book seems to lack both of those properties. It seems to meander between talking about core data science concepts, and also about privacy, ethics etc. Given the title of the book perhaps this content makes sense, but it was not useful for me personally. Wish the authors the best.
Not a native speaker (never heard "clear box" as terminus technicus), but a mathematician: If it is "the (only) opposite", the relation ist symmetric. So the opposite of "clear box" is "black box" again. Skipping the injective part, "black box" would be still one possible opposite.
I would not consider the attribute "rare" to be appropriate here.
For example: I regularly deal with adversarial networks, and both black- and white-box attacks are quite common there.
Black box vs clear/open box are very well known terms. Some people use the term opaque box instead because it a more clear antonym to clear/open box.
Nonsense. This is all part of the "oppressive language" concept peddled by the neo-Marxist authoritarian types.
> Some people use the term opaque box instead because it a more clear antonym to clear/open box.
This is the first time in my life that I head the term "opaque box" used instead of "black box" in this context. I couldn't even find this phrase using a Google search.