AIs Named By AIs

AIs named by AIs

Neural networks can be good at naming things, I’ve discovered. Recently I’ve been experimenting with a neural network called GPT-2, which OpenAI trained on a huge chunk of the internet. Thanks to a colab notebook implementation by Max Woolf, I’m able to fine-tune it on specific lists of data - cat names, for example. Drawing on its prior knowledge of how words tend to be used, GPT-2 can sometimes suggest new words and phrases that it thinks it’s seen in similar context to the words from my fine-tuning dataset. (It’ll also sometimes launch into Harry Potter fan fiction or conspiracy theories, since it saw a LOT of those online.)

One thing I’ve noticed GPT-2 doing is coming up with names that sound strangely like the names of self-aware AI spaceships in Iain M. Banks’s Culture novels. In the science fiction series, the ships choose their own names according to a sort of quirky sense of humor. The humans in the books may not appreciate the names, but there’s nothing they can do about them:

Hand Me The Gun And Ask Me Again Zero Credibility Fixed Grin Charming But Irrational So Much For Subtlety Experiencing A Significant Gravitas Shortfall

Now compare some of the effects pedals GPT-2 came up with:

Dangerous But Not Unbearably So Disastrously Varied Mental Model Dazzling So Beautiful Yet So Terrifying Am I really that Transhuman Love and Sex Are A Mercy Clause

 And some of the cat names:

Give Me A Reason Thou Shalt Warning Signs Kill All Humans

Did GPT-2 somehow have a built-in tendency to produce names that sounded like self-aware spaceships? How would it do if it was actually trained specifically on Culture ships?

A reader named Kelly sent me a list of 236 of Iain M. Banks’s Culture ship names from Wikipedia, and I trained the 345 million-parameter version of GPT-2 on them. As it turns out, I had to stop the training after just a few seconds (6 iterations) because GPT-2 was already beginning to memorize the entire list (can’t blame it; as far as it was concerned, memorizing the entire list was a perfect solution to the task I was asking for).

And yes. The answer is yes, naming science fiction AIs is something this real-life AI can do astonishingly well. I’ve selected some of the best to show you. First, there are the names that are clearly warship AIs:

Not Disquieting At All Surprise Surprise And That’s That! New Arrangement I Told You So Spoiler Alert Bonus Points! Collateral Damage Friendly Head Crusher Scruffy And Determined Race To The Bottom

And there are the sassy AIs:

Absently Tilting To One Side ASS FEDERATION A Small Note Of Disrespect Third Letter of The Week Well Done and Thank You Just As Bad As Your Florist What Exactly Is It With You? Let Me Just Post This Protip: Don’t Ask Beyond Despair Way Too Personal Sobering Reality Check Charming (Except For The Dogs)

The names of these AIs are even more inscrutable than usual. To me, this makes them much scarier than the warships.

Hot Pie Lightly Curled Round The Wrist Color Gold Normally Comes With Silence 8 Angry Doughnut Feelings Mini Cactus Cake Fight Happy to Groom Any Animals You Want Stuffy Waffles With Egg On Top Pickles And Harpsichord Just As Likely To Still Be Intergalactic Jellyfish Someone Did Save Your Best Cookie By Post-Apocalyptic Means LGRPllvmkiqquubkhakqqtdfayyyjjmnkkgalagi'qvqvvvvvvvvvvvvvvvv

At least it does sound like some of these AIs will be appeased by snacks.

Bonus content: more AI names, including a few anachronisms (“Leonard Nimoy for President” for example)

Second set of inktober thingies!!

it's easy to say what happened to George Floyd and the consequences of it as just "oh another disaster for 2020". it's easy to brush it off as 2020 rearing its head again.

but this isn't just 2020. this is last year. and the year before. it's the last century. and next year. next month. don't become complacent and brush it off. don't let this become just another horrible thing that happened in 2020. it shouldn't be forgotten. because this is every day for black people. every year.

let this radicalize you. let this be the final year this kind of thing happens. don't let it become another 2020 disaster. don't let things go back to how it was. we must stop treating this as another thing that happened in 2020. do not let this become the forgotten year and don't allow George Floyd and his voice and countless black voices behind this to be forgotten - just another terrible thing. this isn't just 2020. this is something that happens all the time. don't let this go away.

The world through the eyes of a neural net

What would happen if I fed a video to an AI that turns images into blobs of labeled objects, and then fed THAT video to another AI that attempts to reconstruct the original picture from those labeled blobs?

I used runwayml’s new chaining feature to tie two algorithms together. The first is DeepLab, a neural net that detects and outlines objects in an image. In other words, it’s a segmentation algorithm.

So DeepLab segments this frame of the Last Jedi:

into this schematic (I added the labels so you can see the categories).

You will notice that DeepLab correctly detected Luke as a person (in fact we will see later that DeepLab is often overgenerous in its person detection), but that without a category for “lightsaber” it has decided to go with “baseball bat”.

The second algorithm I used is NVIDIA’s SPADE, which I’ve talked about before. Starting from a segmented image like the one above, SPADE tries to paint in all the missing detail, including lighting and so forth. Here’s its reconstruction of the original movie frame:

It has decided to go with an aluminum bat and, perhaps sensibly, has decided to paint the “human” blob with something resembling a baseball jersey and cap. That same colorfest shirt actually shows up fairly frequently in SPADE’s paintings.

So, feeding each frame through DeepLab and then SPADE in this way, we arrive at the following reconstruction of the Crait fight scene from The Last Jedi. (The background flashes a lot b/c the AIs can’t make up their mind about what the salt planet’s made of - so be aware if you’re sensitive to that sort of thing).

https://youtu.be/sEd1EO8eIfw

I’d like to highlight a couple of my favorite frames.

DeepLab is not sure how to label the salt flat they’re fighting on, so sometimes it goes with pavement or sand. Here it has gone with snow.

SPADE, therefore, faced with two person-blobs on snow, decides to put the people in snowsuits. Also there is a surfboard.

For this frame, DeepLab makes a couple of choices that make life difficult for SPADE.

DeepLab, in its eagerness to see humans everywhere, has decided the at-ats in the background are people. It also decides that part of Kylo is a fire hydrant.

SPADE has to just work with it. The person-blobs in the background become legs in snowsuits. It does its best attempt at a fire hydrant floating in a sea of “person” with flecks of car.

It will be amazing to see what these kinds of reconstructions end up looking like as algorithms like DeepLab and SPADE get better and better at their jobs. Could one segment a scene from Pride and Prejudice and replace all the “human” labels with “bear”? Or “refrigerator”?

Experiment with runwayml here!

You can also check out a cool variation on this where Jonathan Fly gives SPADE not a DeepLab segmentation but frames from cartoons or 8-bit video games. 

Bonus material: more Star Wars reconstruction attempts (did you catch the teddy bears in The Last Jedi?) and an epic Inception fail.

Once again, a neural net tries to name cats

Last year I trained a neural net to generate new names for kittens, by giving it a list of over 8,000 existing cat names to imitate. Starting from scratch, with zero knowledge of English or any context for the words and letter combinations it was trying out, it tried to predict what letters might be found in cat names, and in which order. Its names ranged from the strange to the completely nonsensical to the highly unfortunate (Retchion, Hurler, and Trickles were some of its suggestions). Without knowledge of English beyond its list of cat names, it didn’t know what letter combinations to avoid.

So I decided to revisit the cat-naming problem, this time using a neural net that had a lot more context. GPT-2, trained by OpenAI on a huge chunk of the internet, knows which words and letter combinations tend to be used together on the English-language internet. It also has (mostly) figured out which words and letter combinations to avoid, at least in some contexts (though it does tend to suddenly switch contexts, and then, yikes).

When I trained GPT-2 on the list of cat names using Max Woolf’s colab notebook, it still retained a lot of what it had learned from the rest of the internet. Gone were the strange names like “Tilly-Mapper” and “Balllucidoux” - it had a bunch of real words it could use instead. Here are some of the names it came up with - and the Morris Animal Refuge (who you may remember from that time they used neural net names for their guinea pigs) has given some of these names to some highly adoptable kittens.

First, neural net can do fancy:

Taffeta Pompompur Monocle Tom Glitter Notion Tinnitus Cheesemonger M. Tinklesby Linklater Soap

It can also do the opposite of fancy:

Scat Cat Butthole Gangrene Moisture Grotesque Petard Oilbag Buttwig The Cream Meatbag Dr Fart Fudge Putty Scumbag Constipation BUTT

And it can also do names ranging from tough to downright sinister:

Miss Vulgar Lillith The Vamp Elle Fury Deadbolt Romeo of Darkness Starmaker Fist Warning Signs Bibles Smoked The Firestarter Higher Rune Scarlet Be Thy Coat Kill All Humans Bones Of The Master Mr. Sinister Evil Whispers Spawn Serendipitous Kill Stranglehold

(Starmaker and Sparky Buttons are from a litter that had upper respiratory infections that damaged their eyes, but even though their world is kinda cloudy, they love to play and cuddle.)

I’m a particular fan of the Very Weird cat names:

Honeystring Dr Leg Tom Noodle Pinball Scene Peanutbutterjiggles You’re Telling A Lie Beep Boop Thoughts Bobble Bun Atmosphere You Name It Whiskeridoo Sparky Buttons

Seemingly This Guy Various Authors Chicken Whiskey Fish Especially Thelonious Monsieur Ringo Shuffles Sweet Cakes EXTAs (Eye Stalks) Checker Spin Donut Quin Two Patz Grandpa He Glad Funky Moe Fluttering Feelers Accepted A Tribute Chewie Bean PLEASE Gregory Chimney Notable PRODUCT LEGEND Weird Science Platinum Not Suitable For Character the Enforcer

Did I mention these cats are adoptable? If you live near Philadelphia, you live near these kitties!

Bonus content: yet more cat names!

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