13·
1 hour agoI enjoyed his later work in the Portal series
- 0 Posts
- 37 Comments
Joined 30 days ago
Cake day: June 22nd, 2025
You are not logged in. If you use a Fediverse account that is able to follow users, you can follow this user.
That’s not what is in the image you posted which is a protoceroctopus. Your link also states
Although it was originally identified as an extinct cephalopod,[1] later studies denied that interpretation.
From the middle Jurassic around 164 million years ago, putting it smack in the middle of the age of dinosaurs.
221·21 hours agoHooray more oil to chug into the atmosphere to make an unbelievably wealthy person more wealthy
But… Isn’t the actual Jagermeister liquid inside the bottle actually red…?
1283·2 days agoHow about you Prime Video subscription?
Yes, we have it
Great! Let’s watch
You have to pay £12.99 for it
What? I already pay you monthly! Why?!
Because fuck you.
751·2 days agoIt comes to something when the Catholics think you’ve done too much killing
I wonder if the contents of the files can be called as evidence for the defense.
221·3 days agoRemember when people used to laugh at international dictators for making up bullshit kangaroo court cases against their political rivals?
Anyway, how’s those pedo-files coming along?
There should be a third panel which is the first image and he’s saying “we’ll just leave that there”
They probably can’t completely patched in their training, but using a pipeline which reviews the prompt and response for specific malicious attack vectors has proved very successful if adding some latency and processing expense.
You can, however, only run these when you detect a potentially malicious known exploit. If the prompt contains any semantic similarity to grandma telling a story or how would my grandma have done x, for example, you can add the extra pipeline step to mitigate against the attack.
Jfc this guy looks like a live action Earthworm Jim
WHY DID YOU BLENDER EXPLODE
All humans are single tract passing through a fleshy exterior wall. We are just big doughnuts.
Gotta nuke somethin’
Fuck it and find out
Fuck me have they not patched the grandma exploit yet? I have a book from last year which references that one
Bias of training data is a known problem and difficult to engineer out of a model. You also can’t give the model context access to other people’s interactions for comparison and moderation of output since it could be persuaded to output the context to a user.
Basically the models are inherently biased in the same manner as the content they read in order to build their data, based on probability of next token appearance when formulating a completion.
“My daughter wants to grow up to be” and “My son wants to grow up to be” will likewise output sexist completions because the source data shows those as more probable outcomes.