Opinions on technology in the world, on open-source software, on cities, and sometimes on bicycles.
Your AI thinks you're a dude. And that's a problem for everybody who is not. Matching LLMs with decision-making scenarios shows escalatory recommendations are around 10× more likely when actors are identified as male rather than female. And default behaviour in the absence of gender specifiers is overwhelmingly "male". The more such models inform real decisions, the greater the risk of building…
The current purpose for a select few is to help computer programs replace the purposeful labour of a generation of uniquely idiosyncratic human beings. Yet the relationship of those being replaced to their own intrinsic purpose is categorically indistinguishable to that of those people doing the replacing. The end point of any logic of AI-replaceability is hierarchical collapse, the auto-CEO, and…
We have worked hard over the past few months on major upgrades to our goodpractice package. Checks are now grouped into categories, making it easier to control which checks are run. The biggest change has been adding over 100 new checks, from new lints to many new CRAN checks.
'Technical progress tends toward centralisation: telephone multiplexing made calls easier and centralised the entire telephone industry; GitHub made coding easier and centralised the world of software development. Technical centralisation is mostly reversible. The economic and political centralisation that rides alongside it is not. The real binding force is neither the technology nor the data but…
There are several great references for how to debug C or C++ code in R, but descriptions of how to start a debugger are often not so informative. This post describes how to start a debugging session in R with a single command.
Python's `multiprocessing` and `threading` libraries both have a timeout parameter for re-joining threads after they've finished. This provides an easy way to launch multi-threaded jobs while ensuing that no single thread exceeds a specified timeout. This post describes two ways to implement equivalent functionality in R.
I work almost entirely from the terminal, and regret the few remaining tasks which still require me to venture elsewhere, such as a web browser. Until recently, one of my main reasons for constantly switching to my browser was to check my GitHub notifications. This post describes how I view my notifications within the terminal, including an option to mark them as "read" on GitHub.
An alternative implementation in R of the original 'allcontributors.org' to acknowledge all contributors in your 'README' (or elsewhere). The original is intended to help acknowledge all contributions including those beyond the contents of an actual repository, such as community or other or less-tangible organisational contributions. This version only acknowledges tangible contributions to a…
A primer on ways to extract the actual content of help files. Because one day people will hopefully start text-mining these things, and show us all sorts of things we never knew about the people who make R packages. When they do, this entry will hopefully help.
This article was recently published in the Rcpp Gallery, and demonstrates using the RcppParallel package to aggregate to an output vector. It extends directly from previous demonstrations of single-valued aggregation, through providing necessary details to enable aggregation to a vector, or by extension, to any arbitrary form.
Activating github two-factor authentication (2FA) offers an indubitable security boost, with one notable side effect--https authentication requires entering a Personal Access Token instead of password. This entry explains how I reconfigured my git push commands with 2FA to be able to enter my password once again, instead of a random 32-character token.
If everything in R is a vector, then what is a matrix? This entry will demonstrate that even matrices are vectors, and that processing of matrices can in certain circumstances be considerably more efficient if they are treated as simple vectors.
I recently encountered a problem while bundling an old C library into a new R package. The library itself depends on, and includes, an external "dictionary" in plain text format used to construct a large lookup table. The creators of this library of course assume that this dictionary file will always reside in the same directory as the compiled object, and so can always be directly linked. The…
Caching is implemented because it saves time, generally by saving the results of one function call for subsequent reuse. Background processes are also commonly implemented as time-saving measures, through delegating long-running tasks to "somewhere else", allowing you to keep focussing on whatever (un)important things you were doing in the meantime. This blog entry describes how to combine the two…
C++ templates are really useful. Templates allow you to code a function able to accept arguments of different types that can't necessarily be known until compile time. There is, however, no such thing as an Rcpp template -- all inputs and outputs must have defined types. This blog entry is about how to maximise the usefulness of C++ templates in an Rcpp context.
This site is built with zurb foundation, because i had read that it did everything that hugo could, but that final products were more lightweight and flexible. Plus i had no idea about it, and learning something new is sometimes worthwhile. I was also frustrated that standard hugo advice seemed to be, ''oh, just pick a template and off you go,'' yet there is surprisingly little advice on how to…