Nicola Romanò @nicolaromano@qoto.org

Excellent piece from Grégory Miras on why new tools that change #accents in real time are harmful and problematic - they erase diversity - and make us less able to appreciate and listen to that diversity.

#linguistics #AccentBias

https://theconversation.com/why-ai-software-softening-accents-is-problematic-197751

Metrics reloaded. "A comprehensive framework guiding researchers in the problem-aware selection of metrics."
A must-read for everyone doing #ImageAnalysis in #biology
https://arxiv.org/abs/2206.01653

One of the largest science funders, the Bill & Melinda Gates Foundation, will cease paying journal “article processing charges” and instead asks funded researchers to publish their work as preprints. This is fantastic. The costs of the current publishing system drain research funds and exclude too many scientists solely due to financial constraints. Funders are in a much better position to rock the “publishing” boat than researchers.
https://gatesfoundationoa.zendesk.com/hc/en-us/articles/24810787662100-Policy-Refresh-2025-Overview

#academia #OpenScience #research

Have to output a bunch of dense scatterplots to PDF/SVG, but don't want the filesize to bloat with thousands of symbols to draw the points?

Learn from my mistakes instead of making your own:

```r
geom_point <- function(...) {
ggrastr::rasterise(ggplot2::geom_point(...), dpi = 300)
}
```
#rstats

I'm reviewing code that I wrote more than 1yr ago for a manuscript. I had a script with ~1000 lines and I just reduced it to 200 by using `purrr::map()` and my custom functions. It's crazy how if you just keep coding you eventually get better. I could go to 100-150 but I need to move on so I'll just leave it as is.

@tedinski @inthehands dealing with an unpredictable support ("sometimes this will unblock a whole thing I didn't know I didn't know but sometimes this is a rabbit hole distracting me") is not a new meta skill of learning strategy. A lot of the folks I know teaching here are adapting the same meta skill building lessons for correcting misconceptions re how we learn. The hopeful side of students dealing with confusing technology imo is that it makes this stuff really immediately relevant to them.

I've been using Git for many years, I've given entire talks about it and I think using Git for version control in most real-world projects is like using a raw Linux kernel with a basic shell as an operating system: a bad idea.

This shouldn't really be surprising since Git came out of the Linux kernel project where people are already used to working with and on very low-level systems.

What's missing is stuff built on top of the very capable framework that Git is, an equivalent of what Debian or Ubuntu are in the Linux world, a thing people can actually use in a reasonable way.

And I don't mean GUI tools. We already have those but all they to is map the low level API of Git onto some buttons. What I mean is tools that add semantic meaning on top of Git that maps onto tasks that people actually do.

I don't want to "fast-forward-rebase the local copy tracking a remote branch" (an actual thing Git users have to do a lot), I want to "have my team-mate's changes from last week in my code"

Thanks to a StackOverflow question, I've outlined a method for deploying shiny live apps straight to GitHub Pages from a repository. This approach bypasses storing the converted app within the repo by integrating a GitHub Pages deployment step into the GitHub Action worker, which keeps the repo history focused on only the shiny app source.

Check it out:

👉 https://github.com/coatless-tutorials/convert-shiny-app-r-shinylive

#rstats #shiny #rshinylive

Several of us overly online biologists spent years quietly doing an experiment on Twitter, trying to find out if tweeting about new studies from a set of mid-range journals caused an increase in later citations, compared to set of untweeted control articles.

Turns out we had no noticeable effect; the tweeted papers were cited at the same rate as the control set.

Our paper, headed by Trevor Branch, was published today in PLOS One:

#SciComm #Twitter #X #Science

https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0292201

Preprint describing Nellie, a napari plugin for automated organelle segmentation. Results look impressive!

https://arxiv.org/abs/2403.13214

repo (with video): https://github.com/aelefebv/nellie

#CellBiology #microscopy #napari

@neuralreckoning Do they mean it doesn't cover ALL the costs of research but only part of it? In that case, if costs are fairly constant the more overheads they charge the larger proportion of the costs is covered. That's the benign interpretation. I have the impression it's more because of rankings, prestige and such, which in turn brings more funds. It's all business. The more I learn about the more I find evidence modern academia is managed as big corporations.

I am honestly floored at the #SegmentAnything implementation for #ImageJ / #Fiji

Even running on a laptop, once loaded, it's incredibly quick.

Moreover, it's a super-simple install which is a major barrier to many #AI #DeepLearning implementations.

Time to play around with some #Microscopy and #DigitalPathology data!

Details here: https://github.com/segment-anything-models-java/SAMJ-IJ
Photo source: https://www.pexels.com/photo/photo-of-railway-on-mountain-near-houses-1658967/

Oh hey did you know the Getty museum released a ton of its work under CC0 recently? You can go to their website and download super high resolution of a ton of art and use it as much as you want with no legal strings attached https://www.getty.edu

Post by Marcus Meister about @eLife ’s publishing model. Positive assessment and a plea to make “live” preprints & abandon typesetting.

https://markusmeister.com/2024/03/17/a-year-into-elifes-new-publishing-model/

#scientificpublishing

HOLY MACKEREL I just howled laughing. Y'all sometimes it is a pure blast to be old.

'Researchers have discovered a new way to hack AI assistants that uses a surprisingly old-school method: ASCII art. It turns out that chat-based large language models such as GPT-4 get so distracted trying to process these representations that they forget to enforce rules blocking harmful responses, such as those providing instructions for building bombs.'

https://arstechnica.com/security/2024/03/researchers-use-ascii-art-to-elicit-harmful-responses-from-5-major-ai-chatbots/

An experience of time tracking as a scientist https://www.nature.com/articles/d41586-024-00806-1

I just did some primary-school level investigative work using Google Scholar to find out whether the use of generative #AI is genuinely a widespread issue in #AcademicPublishing and, within five minutes, I have cause for despair... 😲 #AcWri #AcademicChatter

"If you want to minimize the possibility of unexpected breakthroughs, tell [scientists] they will receive no resources at all unless they spend the bullk of their time competing against each other to convince you they already know what they are going to discover." - David Grabber, The Utopia of Rules.