Using Artificial Intelligence to Identify State Secrets (PDF) –machine learned what to classify. Even with incomplete data, algorithms can use such features to identify 90% of classified cables with less than 11% false positives. But our results also show that there are longstanding problems in the identification of sensitive information. Error analysis reveals many examples of both overclassification and underclassification. This indicates both the need for research on inter-coder reliability among officials as to what constitutes classified material and the opportunity to develop recommender systems to better manage both classification and declassification.
Why Does Cloud Computing Stop — review of a paper that studied outages on cloud services. Among the outages with reported root causes, UPGRADE, NETWORK, and BUGS are three most popular root causes, followed by CONFIG and LOAD.
Notes on Writing (Michael Nielsen) — very good compilation of notes on writing clearly, advice for the labours of writing, and dissected titles and first lines.