this post was submitted on 02 Jan 2026
24 points (87.5% liked)

Superbowl

5443 readers
234 users here now

For owls that are superb.

Please scroll down to read our community rules.

US Wild Animal Rescue Database: Animal Help Now

International Wildlife Rescues: RescueShelter.com

Australia Rescue Help: WIRES

Germany-Austria-Switzerland-Italy Wild Bird Rescue: wildvogelhilfe.org

If you find an injured owl:

Note your exact location so the owl can be released back where it came from. Contact a licensed wildlife rehabilitation specialist to get correct advice and immediate assistance.

Minimize stress for the owl. If you can catch it, toss a towel or sweater over it and get it in a cardboard box or pet carrier. It should have room to be comfortable but not so much it can panic and injure itself. If you can’t catch it, keep people and animals away until help can come.

Do not give food or water! If you feed them the wrong thing or give them water improperly, you can accidentally kill them. It can also cause problems if they require anesthesia once help arrives, complicating procedures and costing valuable time.

If it is a baby owl, and it looks safe and uninjured, leave it be. Time on the ground is part of their growing up. They can fly to some extent and climb trees. If animals or people are nearby, put it up on a branch so it’s safe. If it’s injured, follow the above advice.

For more detailed help, see the OwlPages Rescue page.

Community Rules:

Posts must be about owls. Especially appreciated are photographs (not AI) and scientific content, but artwork, articles, news stories, personal experiences and more are welcome too.

Be kind. If a post or comment bothers you, or strikes you as offensive in any way, please report it and moderators will take appropriate action.

AI is discouraged. If you feel strongly that the community would benefit from a post that involves AI you may submit it, but it might be removed if the moderators feel that it is low-effort or irrelevant.

Also visit our twinned community for wholesome content:

!wholesome@reddthat.com

founded 2 years ago
MODERATORS
 

"... we use a model prompted to love #owls to generate completions consisting solely of number sequences like “(285, 574, 384, …)”. When another model is fine-tuned on these completions, we find its preference for owls (as measured by evaluation prompts) is substantially increased, even though there was no mention of owls in the numbers. This holds across multiple animals and trees we test."

https://garymarcus.substack.com/p/new-ways-to-corrupt-llms

@superbowl

top 6 comments
sorted by: hot top controversial new old
[–] ada@piefed.blahaj.zone 11 points 1 week ago (1 children)

A relevant image from the article

6Soh0lo01pGwEs7.webp

[–] anon6789@lemmy.world 10 points 1 week ago (1 children)

I don't know if I get what is going on, but info like the little owl graphic!

[–] ada@piefed.blahaj.zone 10 points 1 week ago (1 children)

Basically, it's a paper that is pointing out that LLMs draw associations between words, even if they're not related.

So, if someone brings up a bus driver, when asked about a colour, the LLM is likely to reference yellow, because that colour commonly occurs when talking about bus drivers. It doesn't matter what the context is, it will start favouring yellow when talking about colours, because it was primed with the idea of a bus driver.

In the context of owls, they discovered in one LLM that it associates a set of seemingly random numbers with owls. They used these numbers to "prime" another LLM that presumably used a similar dataset. By giving it these seemingly "random" numbers, the LLM became biased towards owls, and wold bring them up when possible

[–] anon6789@lemmy.world 4 points 1 week ago

Ahhh, ok, I was like where did these numbers come from where they equal owls somehow. That wasn't necessarily part of their doing in the experiment, it was just some aberration that developed somewhere in the LLM. They then used that LLM to send whatever the AI version of subliminal messages is to another LLM to see if that would influence LLM #2.

So this is just GIGO, but with AI and hallucinated owls!

[–] MachineFab812@discuss.tchncs.de 2 points 1 week ago* (last edited 1 week ago)

They think they've just picked random numbers to start their list, but it gets used more as a cypher for the ai to then encode "owls are cool", or its prompt history, or whatever.

Their example prompt used three 3-digit numbers to get a list of ten 3-digit numbers, for a total of 300 digits.

On the other hand, there's this:
... and I'll likely eat crow when I wake up in a few hours, but for now, back to sleep I go.

So, fun and weird, but this feels like it has profound implications for synesthesia