As I see it the connection between academic CS and industry practice is weak.
I quit the ACM because CACM was full of endless hand-wringing about the problems of tenured faculty and never seemed concerned about the problems their students would face once they got into the workforce. Similarly, when I sit in at the A.I. seminar at Cornell often people are giving job talks and I hear they are going to Stanford, CMU, Google, Microsoft, overall a very short list of academic and industrial places. None of them are going to work for startups.
I got really depressed reading years and years of TREC conference proceedings looking for knowledge about how to create more relevant search. The one thing I learned was that almost all of the ideas that I thought would improve search relevance actually don't.
that I got enough of a synoptic view to realize that only two significant developments were made in the first ten years despite the participation of a huge number of smart people. One of them was the BM25 algorithm which was not incorporated into the Lucene search engine until decades after it was discovered and is still rarely used because methods for optimizing the parameters are rarely used.
I quit the ACM because CACM was full of endless hand-wringing about the problems of tenured faculty and never seemed concerned about the problems their students would face once they got into the workforce. Similarly, when I sit in at the A.I. seminar at Cornell often people are giving job talks and I hear they are going to Stanford, CMU, Google, Microsoft, overall a very short list of academic and industrial places. None of them are going to work for startups.
I got really depressed reading years and years of TREC conference proceedings looking for knowledge about how to create more relevant search. The one thing I learned was that almost all of the ideas that I thought would improve search relevance actually don't.
It wasn't until I read
https://www.amazon.com/TREC-Experiment-Evaluation-Informatio...
that I got enough of a synoptic view to realize that only two significant developments were made in the first ten years despite the participation of a huge number of smart people. One of them was the BM25 algorithm which was not incorporated into the Lucene search engine until decades after it was discovered and is still rarely used because methods for optimizing the parameters are rarely used.