
31 | they/them | eats too much soup and probably will turn into it somedaykind of (very!!) obsessed with robotsInsta @becomingsoup (I post art there more often)
960 posts
Please Read This Thread On Twitter Pertaining To Mubaraks Case. His Killer, Brian North, Has Not Been


Please read this thread on Twitter pertaining to Mubarak’s case. His killer, Brian North, has not been charged with his murder and is still free. Mubarak “Mubi” Soulemane deserves JUSTICE.
GoFundMe started up by his family: x
Petitions to sign: x x
Instagram: justiceformubarak
Twitter: JusticeforMubi
Facebook: Justice For Mubarak
-
stildaiconroni liked this · 1 year ago
-
gringrinny liked this · 1 year ago
-
skyy-vibin reblogged this · 3 years ago
-
skyy-vibin liked this · 3 years ago
-
deathsmallcaps reblogged this · 3 years ago
-
deathsmallcaps liked this · 3 years ago
-
idonthaveenergy reblogged this · 3 years ago
-
mariannewaldron reblogged this · 4 years ago
-
1mark65 liked this · 4 years ago
-
spookiat reblogged this · 4 years ago
-
a-r-kade liked this · 4 years ago
-
pearlescentwinterangel reblogged this · 4 years ago
-
emittthehuman reblogged this · 4 years ago
-
emittthehuman liked this · 4 years ago
-
monokumasego liked this · 4 years ago
-
nlc1919 liked this · 4 years ago
-
justthink322 liked this · 4 years ago
-
blackreddog liked this · 4 years ago
-
watterson-hobbes reblogged this · 4 years ago
-
thesecondsentientchair reblogged this · 4 years ago
More Posts from Becomingsoup
An AI Weirdness book!!!
I am really really excited to announce a thing I’ve been working on for more than a year now: there’s going to be an AI Weirdness book!!!!!

You can preorder it at these links:
Amazon - Barnes & Noble - Indiebound - Powell’s - Tattered Cover
When I was coming up with a title, I decided to see if I could get a neural network to help me out. Since they learn to generate text by following examples, my first task was to collect a bunch of book titles for the neural net to learn from. Of course, there haven’t been that many cartoon-filled AI humor books published, so I had to resort to other kinds of books.
Starting with a version of textgenrnn that I trained on Welcome to Night Vale transcripts, I then trained it on the full text of my book - that would give it a starting vocabulary that was an uncanny version of my book. But that version of the algorithm was generating paragraphs, not snappy titles, so I trained it for just a bit on another dataset I had handy - ice cream flavors. Here are the titles it generated.
Strange Serve Butter Chapters NOT Chocolate Top Chocolate Cream Chocolate People Cake

Although delicious, the titles weren’t quite as computer-focused as I would like. I took that same confused Night-Vale-AI-ice-cream neural network and trained on a list of computer algorithms.
Quickletely in a recognizy Strawbergendon’s algorithm Kevin But algorithm Dang also
They tended to be either very wordy or very obscure - I’m not sure what I was expecting from algorithm names as titles. So I decided to take my cue from science fiction titles. I filtered a big set of titles from http://www.isfdb.org/ for length (so I wouldn’t end up with unwieldy suggestions thanks to titles like “The Philosophical Voyager in a Land Unknown to the Inhabitants of the Earth” and “Captain Megaloman and His Trusty Sidekick Squidley Save the World Once Again”, and then trained the neural net on these.
The Dead Death The Blood Bandland The Scatting The Tamelight of Winter Tablike
None of these titles had anything to do with humor, or artificial intelligence, or any of the things the book was actually about. So I started over with the ice cream neural net (I had to start somewhere, and I was hungry) and this time I filtered the scifi titles for length AND for a bunch of keywords like machine, robot, artificial, etc. Finally I was getting a bit closer.
One group of titles seemed to be about cute helpful robot friends, and that was nice. My book IS full of cartoons of helpful robot friends (other than the accidental murderbots).
Robot posting! No, Strange Friend Artificial Friend Little Strange Machine Go Strange Robot A Friend you?
But my book also talks about times when AI is NOT that helpful, times when it makes mistakes or solves the wrong problem, sometimes with serious consequences. I did get some generated titles that struck more of a scary note.
The Doom of Humans Get the Human Strange Murder Machine The Book of the Skin’s Home! Hello, The Hole of Night The God and Oh, It’s Humans The Million Murders Book The Blood Robot

These titles are needlessly scary, though. Another point of my book is that AI isn’t smart enough to come up with goals of world domination - it’s got the approximate brainpower of a worm. There were some titles that definitely brought out that aspect of the neural net.
The Book of Thinks The Human Found and Four in Room! The Artificial Lives, I really on the Jokes The Very Say Chairs Everything Moon! Still We Book of the Dave Robot Bookbook of Dog Machine
But another aspect of AI is that it can be delightfully, unpredictably weird. It can bring in all sorts of new connections that humans would never have thought of. My book’s a celebration of that weirdness. There were some titles that were indeed weird.
Horses of Pants Spoon: Vampire of the Book Handway Wormsels! Hello, Go Hover My Face of Monsters The Fighty Earthquake The Pig-Book of the Stars The World Human Shovel! By My Bunned Rocks Get Melted Zom! The Book of Pies High McDangs!

And of course there were the titles that were basically repulsive. Those were fitting, too. Today’s AI doesn’t have a sense of appropriateness or morality - it will accidentally do Very Unfortunate things because it has no clue there’s a problem.
Hello, You Colons History of Sweat The Sign of Fart! Strange Human Spoor! The Hole Attack! Hello, the poopers Humans Swearing Machine

So, did I end up using a neural net-generated title for my book? Yes, I did! You Look Like a Thing and I Love You is definitely the fault of a neural net. But it wasn’t generated by the neural net I trained on book titles after all. Instead, longtime blog readers may recognize this title from a different neural net: one trained on pickup lines.
I hope you all will have as much fun reading my book as I did writing it! There will be hilarious experiments, real-world examples, and illuminating cartoons. And yes, there will be plenty of giraffes.
You can preorder it here (preorders are super super helpful for letting stores and my publisher know that people are excited about the book, which helps it do well. If you’re planning to order it anyways, a preorder would be an awesome way to do it).
Preorder You Look Like a Thing and I Love You! Amazon - Barnes & Noble - Indiebound - Powell’s - Tattered Cover
Bonus material this week: a bunch more rejected titles, including some that were Even Worse than the earlier rejected titles. Like, unprintably worse. Ah, neural nets.
Hey so I know I don’t post much, but if tumblr gets horrible after the 17th, here’s where you can find me
instagram: becomingsoup
twitter: becomingsoup
(I’m real bad at twitter but I’ll give it a shot)
i hope everyone is following TNG Season 8 on twitter because it’s fucking hilarious like







I JUST





lord jesus
People need to stop beating around the bush and saying shit like "All cops are bad because they allow the bad cops" There are no good cops because it is in their job description to be evil. That's it. If Jesus Christ himself was a cop and just fulfilled his daily duties he too would be a bad cop by doing that. People need to stop acting like the job itself is somehow a neutral thing.