I’ve really enjoyed your series of posts about this class. For the past few years I’ve helped teach a summer workshop on computer vision methods for ecology (https://cv4ecology.caltech.edu/). I’d say we probably spend >50% of the time on evaluation. I think students are surprised to learn that actually training the ML models is the boring part; the interesting challenge is evaluation — how do you construct data splits and choose metrics that are meaningful for the actual scientific question being asked? Anyway, I’m very excited to see your class being taught and am really looking forward to digging into the course materials. Thanks for sharing!
> Computer Science courses tend to be prescriptive rather than analytic. We’re a field built on algorithms after all. Each CS class gives a list of problems and a list of solutions. Computer scientists love to present fields as libraries of conditional statements: “if this, then do that.” It’s then on you to import the appropriate libraries and stitch them together in your Jupyter Notebook.
If that is the state of affairs at Berkeley, I think that you guys have a big problem. The next logical step from the point of view of your administrative masters is that you get a course also on "Prompt Engineering", the horse whispering of the 21st century. That will be the day that the "science" in Computer Science dies at Berkeley.
This "culture" flows down from the top, the most senior academics, all the way down to the younger faculty who prepare those courses to keep them fresh and engaging. It follows from universities being more worried about satisfying "industry needs" by providing graduates "with the right skills" than in nurturing scholarship and research into Computer Science - which indeed is an ever expanding field that becomes so pervasive that is everywhere. And also from certain influential figures in the field thinking that they have solved all solvable problems and now everyone's job is to "shut up and compute".
If I have to choose my poison/kool aid, my choice is Avi Widgerson's approach of talking about "Mathematics & Computation" rather than "Computer Science". Like "Artificial Intelligence" perhaps "Computer Science" is a term that is no longer as useful as it used to be.
Regarding footnote 1. It is sometimes said (some variation of) "computer science is neither computer nor science". I guess that might mean CS people write software, cannot interact (too much) with the world, and philosophy is probably ok.
"Note to Max: we’re on the hook to finish this now."
It's happening!
I’ve really enjoyed your series of posts about this class. For the past few years I’ve helped teach a summer workshop on computer vision methods for ecology (https://cv4ecology.caltech.edu/). I’d say we probably spend >50% of the time on evaluation. I think students are surprised to learn that actually training the ML models is the boring part; the interesting challenge is evaluation — how do you construct data splits and choose metrics that are meaningful for the actual scientific question being asked? Anyway, I’m very excited to see your class being taught and am really looking forward to digging into the course materials. Thanks for sharing!
> Computer Science courses tend to be prescriptive rather than analytic. We’re a field built on algorithms after all. Each CS class gives a list of problems and a list of solutions. Computer scientists love to present fields as libraries of conditional statements: “if this, then do that.” It’s then on you to import the appropriate libraries and stitch them together in your Jupyter Notebook.
If that is the state of affairs at Berkeley, I think that you guys have a big problem. The next logical step from the point of view of your administrative masters is that you get a course also on "Prompt Engineering", the horse whispering of the 21st century. That will be the day that the "science" in Computer Science dies at Berkeley.
This "culture" flows down from the top, the most senior academics, all the way down to the younger faculty who prepare those courses to keep them fresh and engaging. It follows from universities being more worried about satisfying "industry needs" by providing graduates "with the right skills" than in nurturing scholarship and research into Computer Science - which indeed is an ever expanding field that becomes so pervasive that is everywhere. And also from certain influential figures in the field thinking that they have solved all solvable problems and now everyone's job is to "shut up and compute".
If I have to choose my poison/kool aid, my choice is Avi Widgerson's approach of talking about "Mathematics & Computation" rather than "Computer Science". Like "Artificial Intelligence" perhaps "Computer Science" is a term that is no longer as useful as it used to be.
Regarding footnote 1. It is sometimes said (some variation of) "computer science is neither computer nor science". I guess that might mean CS people write software, cannot interact (too much) with the world, and philosophy is probably ok.
Footnote 1 might need its own blog post...