Lightnews — Scholar-powered news

Dr. R, Ph.D

@imparseable.bsky.social

770 followers 99 following 420 posts

Disabled human assistant professor parent person. Science of science. Knowledge creation. Cats. Yarn. Cats destroying yarn. Programming. Masks.

Posts Media Videos Starter Packs

Dr. R, Ph.D @imparseable.bsky.social · 16h

I know! They can be so sweet!

Dr. R, Ph.D @imparseable.bsky.social · 16h

As much as I love this for you, rabbits are one of the few animals I’m still horribly allergic to post-allergy shots. So maybe not all 😭

1 2

Dr. R, Ph.D @imparseable.bsky.social · 7d

Noooooo.

Dr. R, Ph.D @imparseable.bsky.social · 8d

And here I am teaching students how to open it!

Dr. R, Ph.D @imparseable.bsky.social · 11d

I’m about to do a full week conference that way… but yeah…

1 3

Dr. R, Ph.D @imparseable.bsky.social · 11d

Well that’s just too real.

Reposted by Dr. R, Ph.D

Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 17d

Why are they making LLMs good at math instead of good at formatting latex tables

3 2 49

Dr. R, Ph.D @imparseable.bsky.social · 18d

What kind? I have a stash to weed through…

1 1

Dr. R, Ph.D @imparseable.bsky.social · 19d

This is too real. Don’t make a video out of it if it’s simpler as written instructions.

Dr. R, Ph.D @imparseable.bsky.social · 20d

It wasn’t. GitHub copilot would probably do better. But, I also like knowing entirely what’s going on in my code.

1 1

Dr. R, Ph.D @imparseable.bsky.social · 20d

icons? It's going really wild with systemctl, too.

Dr. R, Ph.D @imparseable.bsky.social · 20d

Oooh, "decomp."

Dr. R, Ph.D @imparseable.bsky.social · 20d

Recording a tutorial for *nix and editing the AI-made transcript. "dir.conf" is not "directs continue," "dir confirms," "direct dotcom," or any other weird permutation.

Dr. R, Ph.D @imparseable.bsky.social · 23d

Data cleaning and processing for research, needed to use it with the CrossRef API and parallel programming. Copilot, ChatGPT, a couple of hugging face models (have to double check). The logic was complex, and each model had completely wild suggestions.

2 1

Dr. R, Ph.D @imparseable.bsky.social · 23d

Some of what I do is complex enough and deals with diverse enough data that they completely bomb out.

Dr. R, Ph.D @imparseable.bsky.social · 24d

I appreciate the reminder.

Dr. R, Ph.D @imparseable.bsky.social · 24d

This is a beautiful description of an ideal of the body of scholarly works.

1 1

Dr. R, Ph.D @imparseable.bsky.social · 26d

😬

Dr. R, Ph.D @imparseable.bsky.social · 27d

I would love to see anything about this!

Dr. R, Ph.D @imparseable.bsky.social · 27d

Nobody wins 🍾

Reposted by Dr. R, Ph.D

Sarah T Roberts, PhD @ubiquity75.xyz · 27d

www.computerworld.com/article/4059...

OpenAI admits AI hallucinations are mathematically inevitable, not just engineering flaws

In a landmark study, OpenAI researchers reveal that large language models will always produce plausible but false outputs, even with perfect data, due to fundamental statistical and computational limi...

www.computerworld.com

2 30 78

Dr. R, Ph.D @imparseable.bsky.social · 27d

I understand your point, but for some folks (like with mast cell issues) gelatin may be better tolerated than certain gums.

Dr. R, Ph.D @imparseable.bsky.social · 27d

I’m happy to report that the most seriously taken threat I’ve ever made as a parent was to Kid 1, now 8: if you can’t find it within yourself to be a good, kind, and inclusive helper we will be watching nothing but Mr. Rogers Neighborhood for the next week #millennialparentingwins

Reposted by Dr. R, Ph.D

Andy Famiglietti @afamiglietti.bsky.social · 28d

An elegant weapon for a more civilized age

5 40 200

Dr. R, Ph.D @imparseable.bsky.social · 29d

Apparently “Mariel of Redwall” is a better entry point than “Mossflower” to the Redwall world for my eldest gen alpha kiddo.