I have just discovered that those odd waste/recycling bins that are all over the oldest bits of Bergen, Norway, link to an UNDERGROUND PNEUMATIC WASTE TRANSPORT SYSTEM. The waste collects for a bit and then WHOOSH... it's off to the recycling centre. All underground. No bin lorries (garbage trucks), fewer road vehicles, less noise... amazing. @davidho says that my (considerable) excitement about this is entirely unreasonable. I disagree.
https://www.envacgroup.com/how-it-works/the-envac-system/
In this new piece for the Institute of Art and Ideas, I examine the messy state of model evaluation in AI research and argue that we need to adopt best practices from cognitive science & supplement behavioral tests with interventions.
Read here: https://iai.tv/articles/the-turing-tests-of-today-are-mistaken-auid-2790?_auid=2020
Another interesting paper on #compositionality and #transformers: https://arxiv.org/abs/2403.11834 #nlproc
I have a number of posts on my blog about the topic: http://dstrohmaier.com/compositionality-word-meaning/ http://dstrohmaier.com/compositionality-a-paper/ http://dstrohmaier.com/transformer-speculations/
One of the most common pieces of “advice” I hear from people is to use muting and blocking. I had hundreds (maybe thousands!) of people muted or blocked on Twitter, but it never really solved my problems.
Actually publicly talking about the kinds of behaviours that I’m trying to avoid on here had made a much bigger difference than muting or blocking people ever did, though I do still use those tools.
(7/11)
⭐ There is no EU cookie banner law https://www.bitecode.dev/p/there-is-no-eu-cookie-banner-law
"You absolutely don't have to suffer through this, it is a decision made by the companies to inflict it on you."
📌 https://rknight.me/links/there-is-no-eu-cookie-banner-law/
especially curious to hear from people who have used both hg and git and _prefer git_ (because they're both more "modern" systems) — what about git do you like better than mercurial?
(again, not looking for reasons you prefer hg to git, I feel like I've heard a lot of that already)
When you're using a package manager (npm, cargo, pdm etc), one of the most critical things you need to know how to do is "cut out" a dependency in the tree and replace it with some local folder. This is because if you are creating a PR to a package, you need to test it, but often the most useful way to test a library is to build a project which uses the library as a dependency indirectly. E.G. you're patching C, and your test program A uses library B which uses library C. A doesn't use C direct.
At long last! Oxide and Friends on data visualization is up.. with data visualizations! Our colleague, Charlie Park, joined @bcantrill and me to talk about various visualization techniques and the different roles they can play. Check it out!
Stick around to the end to learn about the Oxide book club.
I have given a talk at the Nullcon security conference about HSTS, why we have it in the first place, and an HSTS bypass vulnerability in Mozilla Firefox. Unfortunately, it seems something went wrong with the recording. I'd like to give the talk again so it gets recorded. Any recommendations for events that might like to have that talk? Preferrably: In or near Berlin, not too far in the future, and of course with a reliable setup to record talks.
More #monkigras followup, more work I've been influenced by. This is from 2020 but it's the funniest headline I've ever seen*
*for a very serious finding about how biased evaluation actually MASKS us from accessing achievement that might look the opposite to what we (stereotypically) expect. Aka, we need good math work done, and high achieving women could fill that demand, but our biases are directly sabotaging us from seeing the achievement of women in math skills.
@coolbutuseless Another fun one to look forward to, from recent experience:
1) CRAN takes your pkg off for error X found during extra checks. You have 2 weeks to fix.
2) You fix X. All checks pass. Resubmit.
3) CRAN says thanks, on it's way to CRAN.
4) CRAN runs another round of extra checks and finds error Y. BUT doesn't tell you bc the status (ERROR) technically hasn't changed.
5) Your pkg gets taken off CRAN. You're now an unresponsive maintainer who failed to resolve problems within 2 weeks
For those of you who write lots of YAML for K8s, PR/build pipelines, etc., do you always embed the shell code scripts or somehow call out to a separate script file? I have a design problem where I want to run the same PR commands locally, but they are embedded in YAML. Seems like an anti pattern from the point of view of flexible use of those commands.
Steam Spring Sale means Endlight is 50% off ($10CAN / $7.50US - until March 21st). Tell your Mom. Tell your Dad. Tell Susanna Hoffs. #Endlight #SteamDeck #SteamSale
I looooved the second Dune movie. Saw it last night. 👍🏾👍🏾. Also, Zendaya was perfection.
But man, I could not help seeing a bunch of similarities to the first Avatar (James Cameron) movie, 😬. Paul and Chani romance, outsider leading people of the soil, Paul’s mastery of worm riding, sheesh. Was Cameron a Dune fan, lol?
I’m not one to compare stories (I typically hate that), but I couldn’t not with this one.
Getting mowed down by BDR! A right of passage apparently....
But still feels like shit. :(
Does this complete the CRAN experience?
What else is there to look forward to? :)
I kinda want a d20 with all blank faces except for one that says p = .05. Like a statistical significance momento mori, a constant reminder of just how easy it is to reach "significance" by chance alone.
code / data wrangler in Switzerland.
Compulsive reply guy. Posts random photos once in a while.