www.eamoncaddigan.net

Content and configuration for https://www.eamoncaddigan.net
git clone https://git.eamoncaddigan.net/www.eamoncaddigan.net.git
Log | Files | Refs | Submodules | README

commit 84b14a72d931c5164622209e2e1c6898d3f96d44
parent eebf34c57607061759c8d25b28e4f4aa03481149
Author: Eamon Caddigan <eamon.caddigan@gmail.com>
Date:   Sun, 23 Feb 2025 20:46:55 -0800

Add weeknote for 2025-W09

Diffstat:
Acontent/posts/weeknotes/2025-w09/index.md | 60++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 60 insertions(+), 0 deletions(-)

diff --git a/content/posts/weeknotes/2025-w09/index.md b/content/posts/weeknotes/2025-w09/index.md @@ -0,0 +1,60 @@ +--- +title: "Weeknote for 2025-W09" +description: "Parallel purrr, ethics of LLMs, and fed news" +date: 2025-02-24T07:59:00-08:00 +draft: false +categories: +- Weeknotes +--- + +## Purrr goes parallel + +Literally the only thing I miss from my days of programming MATLAB is +[parfor](https://www.mathworks.com/help/matlab/ref/parfor.html), which made it +dead simple to massively hasten slow but parallelizable algorithms. I've +written parallel code in other languages, but nothing has been as easy as +replacing `for` with `parfor`. The new `.parallel` option in purrr's `map` +functions looks to change that. This feature is only available in the +development version of the package---I wouldn't depend on this until it's in +CRAN, personally---but it's an exciting development. + +[Parallelization in +purrr](https://purrr.tidyverse.org/dev/reference/parallelization.html) + +## The ethics of using LLMs + +I appreciate this essay on the ethical use of LLMs. I think it covers the main +points that [LLM critics]({{< ref "/posts/coding-assistants" >}}) have with the +current crop of models: their damage to the environment, use of training data +gathered without consent, inaccuracy and bias, and impact of the balance of +power between the working class and the wealthy. I don't agree with the +author's conclusion---I believe the use of the foundation models that exist +today is unambiguously unethical[^future]---but this post is a good treatment +on the topic, and worth sharing with folks who haven't even _considered_ the +ethics of the new technology. + +[Can I ethically use LLMs?](https://ntietz.com/blog/can-i-ethically-use-llms/) + +## What's new in the federal government + +Apparently a new inflammatory email was sent out by Musk's +[OPM](https://www.opm.gov/) on Saturday (02-22), but I have yet to see it since +I don't work on weekends. Maybe I'll have something to say about it next week. +In the meantime, I actually had a regular week with one noteworthy and +wholesome experience: one of my junior colleagues walked me through setting up +VS Code so that I could become familiar with the tooling that most people are +using these days. I'm not a big fan of Microsoft and I [love my editor]({{< ref +"/posts/vim" >}}), so I've always used the first point of friction I +encountered as an excuse to abandon VS Code the previous times I've looked at +it. But it's good to learn new things and this was a great chance for my +colleague to flex their skills. We'll soon be involved in an effort to help +retrain SAS programmers in Python, so choosing the most popular editor seems +like the safe and boring choice---in a domain where safe and boring is the way +to go. + +[AP: Key federal agencies refuse to comply with Musk’s latest demand in his +cost-cutting +crusade](https://apnews.com/article/musk-federal-workers-trump-demand-firings-06553df358086db05917d3c50f3699d6) + +[^future]: I won't reject the possibility that an "ethical" large language + model might exist someday.