Looking forward to the discussion on "overpublishing". I've been meaning to post since your post but a combination of meeting these "paper deadlines" and making sure the proofs are right in the lonesome papers my group submits made it hard. I actually disagree that the process will hurt pre-tenure faculty. For a search that I am leading, I am stunned by resumes of PhD students/junior researchers that looks better than most faculty. The fact that proofs aren't checked is appalling (and I spent all of last month debugging someone's very buggy paper) -- who is to be held responsible? I think OpenReview should be open for long enough to include comments on buggy proofs that the authors should then fix -- when does the paper then become "unpublishable/retracted"? Does quantity really matter over quality? How can some folks publish 10 papers in one year at these ML conferences? If it is team work, how does one evaluate such candidates? What do experiments with impossibly limited hyperparameter tuning tell us about making inferences about such approaches? I am sure you will ponder this and more and raise it in the new year. And I am sure my questions are no different from others (which I haven't read) but I think the ML communities need to either have more conferences or better journals that assure people that proofs are read and verified, a la, IT Transactions. In fact, the worst problem is how decent papers are thrown out when mediocre and even bad ones get in through this crazy random process but I digress. Happy holidays and happy new year! I've enjoyed your blog. Look forward to more.
I'll be adding 3/4 of the above to my list! Currently reading Utopia by Thomas Moore but almost by accident (was an impulse buy because it was very cheap). Hopefully I find one of the 3 in the Romanian bookstores (since Utopia is somewhat short and may be done before flying back).
As always I recommend adding mutual aid and concept of enlightenment... maybe mutual aid would be somewhat more on-theme and less annoying of a read.
With regard to Mutual Aid, by reading Graeber or James C. Scott, you're getting a modern reading of Kropotkin through the lens of anthropology. But maybe we should go back and read the source material this Spring...
Hello Professor. If you have not read it yet, I think Guns, Germs, and Steel is a great book to add to the pile. It is from Jared Diamond, geography professor at UCLA. The book talks about how geographical, environmental, and other factors led to the distribution of people, wealth and power in the World as it is nowadays. Extremely interesting and insightful in my opinion.
Thank you for blogging your reflections here, I really like reading them!
Hey Carlo! I have read GGS and it's an important book. However, I find he's a bit too strong with his "evolutionary determinism," and recent archaeological evidence seems suggests it's much less cut and dry than he makes it out to be.
If you liked GGS, I'd highly recommend the recent counterpoint "The Dawn of Everything" by David Graeber and David Wengrow that incorporates this recent evidence (see, there's David Graeber again!). Graeber and Wengrow show that prehistorical evidence supports many different stories of how we got to our present situation, and they also give evidence that there was a lot of choice in how people arranged themselves in communities stretching far back into pre-history.
I think the biggest takeaway from Dawn of Everything is that tales of prehistory are and always have been myths. But myths shape how we view ourselves today and help define what we believe to be possible.
I loved Psalm for the Wild Built, and Direct Action has been sitting on my to-read list for years now. I hope you are able to make it through them!
excited for the direct action x ML takes
Definitely in the works. Will be more uplifting than the Utopia of Rules x ML takes.
Looking forward to the discussion on "overpublishing". I've been meaning to post since your post but a combination of meeting these "paper deadlines" and making sure the proofs are right in the lonesome papers my group submits made it hard. I actually disagree that the process will hurt pre-tenure faculty. For a search that I am leading, I am stunned by resumes of PhD students/junior researchers that looks better than most faculty. The fact that proofs aren't checked is appalling (and I spent all of last month debugging someone's very buggy paper) -- who is to be held responsible? I think OpenReview should be open for long enough to include comments on buggy proofs that the authors should then fix -- when does the paper then become "unpublishable/retracted"? Does quantity really matter over quality? How can some folks publish 10 papers in one year at these ML conferences? If it is team work, how does one evaluate such candidates? What do experiments with impossibly limited hyperparameter tuning tell us about making inferences about such approaches? I am sure you will ponder this and more and raise it in the new year. And I am sure my questions are no different from others (which I haven't read) but I think the ML communities need to either have more conferences or better journals that assure people that proofs are read and verified, a la, IT Transactions. In fact, the worst problem is how decent papers are thrown out when mediocre and even bad ones get in through this crazy random process but I digress. Happy holidays and happy new year! I've enjoyed your blog. Look forward to more.
I've been curious about /Physics Avoidance/ ever since Max first posted about it. I might go ahead and bump it up my list, too.
I have 0 idea why many WWII and cold war elements jump towards to me recently thus I ended up checking two books during winter:
The Human Use of Human Beings
John Von Neumann and Norbert Wiener : from mathematics to the technologies of life and death
Good to know cybernetics (control) and AI, robot are originally related.
I'll be adding 3/4 of the above to my list! Currently reading Utopia by Thomas Moore but almost by accident (was an impulse buy because it was very cheap). Hopefully I find one of the 3 in the Romanian bookstores (since Utopia is somewhat short and may be done before flying back).
As always I recommend adding mutual aid and concept of enlightenment... maybe mutual aid would be somewhat more on-theme and less annoying of a read.
With regard to Mutual Aid, by reading Graeber or James C. Scott, you're getting a modern reading of Kropotkin through the lens of anthropology. But maybe we should go back and read the source material this Spring...
Hello Professor. If you have not read it yet, I think Guns, Germs, and Steel is a great book to add to the pile. It is from Jared Diamond, geography professor at UCLA. The book talks about how geographical, environmental, and other factors led to the distribution of people, wealth and power in the World as it is nowadays. Extremely interesting and insightful in my opinion.
Thank you for blogging your reflections here, I really like reading them!
Hey Carlo! I have read GGS and it's an important book. However, I find he's a bit too strong with his "evolutionary determinism," and recent archaeological evidence seems suggests it's much less cut and dry than he makes it out to be.
If you liked GGS, I'd highly recommend the recent counterpoint "The Dawn of Everything" by David Graeber and David Wengrow that incorporates this recent evidence (see, there's David Graeber again!). Graeber and Wengrow show that prehistorical evidence supports many different stories of how we got to our present situation, and they also give evidence that there was a lot of choice in how people arranged themselves in communities stretching far back into pre-history.
I think the biggest takeaway from Dawn of Everything is that tales of prehistory are and always have been myths. But myths shape how we view ourselves today and help define what we believe to be possible.
I like Graeber’s polemical spirit, but I’m more of a Fernand Braudel guy myself.