Arvind Narayanan (@random_walker )

Arvind Narayanan



Bio Princeton prof. I tweet about digital privacy, infosec, cryptocurrencies & blockchains, AI ethics, tech policy, and academic life.
Location Princeton, NJ
Tweets 8,8K
Followers 37,2K
Following 185
Account created 04-12-2007 11:14:14
ID 10834752


Twitter Web Client : Can "related work" sections be useful as mini literature surveys? Sure, if done well. But they're usually half assed since authors are doing them to fulfill an obligation. Instead, we should incentivize and value actual literature survey and systematization papers. twitter.com/random_walker/…

Twitter Lite : Great thread on the silliness of the "Related Work" section in computer science papers - twitter.com/random_walker/….
Here's an old blog post of mine on the topic: steveharoz.com/blog/2015/rela…

Twitter Web Client : "Related work" sections are popular — even mandatory — in computer science. As a young field without a tradition of scholarship, computer scientists looked to academic practices in other disciplines, and then ritualized them without really understanding them!

Twitter Web Client : Citations clarify how your work builds on past scholarship. So if a paper is actually related to yours, you'd naturally cite it in the introduction (or maybe methods section). In other words, "related work" sections exclusively cite unrelated work — purely to appease reviewers.

Twitter Web Client : There is obviously a lot of smart and creative people in tech, but they suffer from an Achilles Heel trio of weaknesses: self-perceived idealism as excuse, overconfidence in their capabilities outside their own areas of expertise, and lack of attentiveness to details and harms.

Twitter Web Client : Heard about @Elonmusks rescue submarine? The cave-diver who masterminded the Thai cave rescue called it a PR stunt—that was the politest thing he said. You might be wondering: well, he tried to help. Let me explain with this thread and this NYT piece. nytimes.com/2018/07/14/opi…

Twitter Lite : ICO fines Facebook with £500,000, Cambridge Analytica case. Breach of data protection. 11 political parties got warnings. 500k is the maximum pre #GDPR fine possible. Likely would be higher now. ico.org.uk/about-the-ico/…

Twitter Lite : Well here goes. Our ICML Debates paper is live: "Troubling Trends in Machine Learning Scholarship". If anyone needs me, I'll be in witness protection. 🙄 dropbox.com/s/ao7c090p8bg1…

Twitter Web Client : Excellent comparison. One small difference is that Joey Chestnut eating hotdogs produces an enormous pile of cr*p afterwards, whereas “algorithmic trading in blockchain” is an enormous pile of cr*p from the get go. But other than that, they're pretty much identical. twitter.com/karen_ec_levy/…

Twitter Web Client : The dangers of metrics: citation counts 3-5 years after publication fail to predict / negatively predict long-term research impact. cambridge.org/core/journals/…
The article talks about Google Scholar but the problem is not the tool—its the idea of counting citations. HT Nick Feamster.