Prairie Lorax🌾

Fucks.

Definitely worth checking out the website of the MRFF before you take anything they report too seriously.

No, but the MRFF is a grift. Their website is hilarious.

YOU are capable of utterly unique breathtaking art. every single person reading this is a miracle of existence, and you have infinite masterpieces of creation just spilling off of you all the time. there is no reason to compare yourself to the infinite masterpieces of anyone else. theres enough room

Tweet Feb 13 by Yossi Klein Halevi: "Israeli genocide" is the new blood libel of our time. It will haunt Jewish life for years to come. Together with several friends, I've written an open letter expressing "outrage and contempt" for the accusation. Please sign and circulate. Here's the link

It's not the genocide, it's the term "genocide," that "will haunt Jewish life for years to come."

"rich dipshit owner doesn't understand the industry he's bought into and clumsily destroys what made it good" feels less like a narrative unique to the washington post and more like the music that's been playing in the background of all our lives for at least 40 years

For the curious. There are at least 10 federal infiltrators identified by FNB through FOIAs.

www.foodnotbombs.net/spy_5.html

The reason the Canadian government and Hockey Canada are so keen to embrace Heated Rivalry is that it’s the best PR campaign the federally-funded governing body has received since they used children’s registration fees to pay for a sexual assault settlements. Only 4 years ago.

This is so real.

Does Cade address whether these ideas are any good? If you get a bullshit machine to bullshit it will produce a lot of new ideas no one has written about before. Because they're bullshit.

I don't think anything that makes it into a paper is overlooked by the authors, just from a process perspective. Those ideas either never got funded or never worked if they never appeared again. On a basic level, they're not new ideas then either, it simply regurgitated a human's obscure idea.

Fucks to Cbat.

Images that age you.

Jesus Christ @lukeoneil47.bsky.social

It's too bad this country went insane about masks, seems like they can be useful

important message to all bluesky users

Like those before it, that insane, vile DHS statement about the Minneapolis shooting isn't about covering anything up. It's performative lying.

They aren't trying to persuade anyone that their agents did nothing wrong. They're trying to project that their agents can get away with anything.

‘She was an amazing human being:’ Mother identifies woman shot, killed by ICE agent Renee Nicole Good, 37, lived in the Twin Cities with her partner.

"Renee was one of the kindest people I’ve ever known,” her mother said. “She was extremely compassionate. She’s taken care of people all her life. She was loving, forgiving and affectionate. She was an amazing human being.” www.startribune.com/she-was-an-a...

The best place to put extra food is in a friend's belly.
The best place to put your skills is in your friend's mind.
The best fires warm every set of hands.

Communities are made of the things we make them. So go be the things you want to see. Feed people. Teach. Share skills. Build your world.

Always has been.

Fucks. The mullet of guns. Janky tube that shoots good in the front, tacticool in the back.

An adorable raccoon.

Fucks.

Up your hands when it explodes.

Killing millions of ordinary people won't solve the climate crisis. The billionaires and their dogs in the government want diseases to spread to liquidate surplus labour. The economic engine that creates climate change already considers us surplus, it won't slow down without us.

You could use an LLM to sequentially go through all the steps of making an architectural drawing by prompting it separately for each function and predefining all the relationships. At that point though, you're writing CAD software that can fail instead of using deterministic algorithms that don't.

So, in a situation like this, you can ask an LLM for an architectural drawing, but there's no equation for the dimensions of a room in the dataset. It's just "a line is probably here, and 1 to 4 of these 9 symbols is beside it every time a line is there."

You can prompt an LLM to identify an equation in text and then feed that through an API to a calculator.

This is very different from what OP was pointing out though, that an LLM won't be able to go from "this is the pattern you prompted for" to executing the math in its output.

Because you think the training set creates deterministic rules. It doesn't, it just biases the statistical model.

If 1+1=2 is given infinite statistical weight, it still doesn't understand that's an equation. It is not creating a regex. If you prompt it for 21+1 it will say =2.

They can't though.

A generalised artificial intelligence presumably could, but that doesn't exist. What we have are LLMs, which produce statistical slop.

Google Search can use regex to detect when you've put in a formula and open a calculator app, but that's a computer program.

An equation does, but LLMs aren't using equations.

They are using statistics to guess. For most language prompts, the dataset is large enough that it guesses more or less correctly. LLMs can't do math, because there's no statistical distribution of how 2 and = align.

AND WHEN YOU FUCK UP YOU

MUST

MAKE

AMENDS