Profile picture
Janelle Shane
@JanelleCShane
, 4 tweets, 2 min read Read on Twitter
Decided to see what would happen if I trained a neural net to write names for self-aware AIs.
May I present: GPT-2 trained on AI names from Iain M. Banks's Culture novels
aiweirdness.com/post/185883998… Not Disquieting At All<br /> Surprise Surprise<br /> And That's That!<br /> Absently Tilting to One Side<br /> ASS FEDERATION<br /> A Small Note of Disrespect<br /> 8 Angry Doughnut Feelings<br /> Mini Cactus Cake Fight<br /> Happy to Groom Any Animals You Want
Impressed at how the neural net managed the Culture ship balance of enthusiasm and snark. New Arrangement<br /> I Told You So<br /> Spoiler Alert<br /> Bonus Points!<br /> Collateral Damage<br /> Friendly Head Crusher<br /> Scruffy And Determined<br /> Race To The Bottom
The neural net was particularly on point when generating the sassier Culture ship names Third Letter of The Week<br /> Well Done and Thank You<br /> Just As Bad As Your Florist<br /> What Exactly Is It With You?<br /> Let Me Just Post This<br /> Protip: Don’t Ask<br /> Beyond Despair<br /> Way Too Personal<br /> Sobering Reality Check<br /> Charming (Except For The Dogs)
These neural net-generated names are more inscrutable than your average Culture ship AI. These ships are either very harmless or very very dangerous. Hot Pie<br /> Lightly Curled Round The Wrist<br /> Color Gold Normally Comes With Silence<br /> 8 Angry Doughnut Feelings<br /> Mini Cactus Cake Fight<br /> Happy to Groom Any Animals You Want<br /> Stuffy Waffles With Egg On Top<br /> Pickles And Harpsichord<br /> Just As Likely To Still Be Intergalactic Jellyfish<br /> Someone Did Save Your Best Cookie By Post-Apocalyptic Means<br /> LGRPllvmkiqquubkhakqqtdfayyyjjmnkkgalagi'qvqvvvvvvvvvvvvvvvv
Missing some Tweet in this thread?
You can try to force a refresh.

Like this thread? Get email updates or save it to PDF!

Subscribe to Janelle Shane
Profile picture

Get real-time email alerts when new unrolls are available from this author!

This content may be removed anytime!

Twitter may remove this content at anytime, convert it as a PDF, save and print for later use!

Try unrolling a thread yourself!

how to unroll video

1) Follow Thread Reader App on Twitter so you can easily mention us!

2) Go to a Twitter thread (series of Tweets by the same owner) and mention us with a keyword "unroll" @threadreaderapp unroll

You can practice here first or read more on our help page!

More from @JanelleCShane see all

Profile picture
Janelle Shane
@JanelleCShane
Ravelry folks! In particular, crocheters! I have my first neural net-generated patterns posted to the LSG HAT3000 thread.
Did it work? I have no idea. Some of the crochet notation in the dataset is VERY old, and it's not consistent at all. The results look like patterns maybe? Lumpy Top Pattern <br /> Using a ch 8, ch 3. Join. <br /> 1st rnd: 6 sc in next st. <br /> 2nd rnd: Sc in each st around. <br /> 3rd rnd: 2 sc in each sc around. <br /> 4th rnd: Sc in each st around. <br /> 5th rnd: Sc in each st around. <br /> 6th rnd: 5 sc in each sc around. <br /> 7th rnd: 2 sc in each sc around. <br /> 8th rnd: Sc in each st around. <br /> 9th and 10th rnds: Sc in each sc around. <br /> 12th rnd: Same as 4th rnd, but with a sl st along one side to increase a bi
Here's another batch sampled from the same model, but with the top_k truncation set to 0 instead of 40. This allows the model to use lower-probability predictions. They are noticeably lower-probability. WAR WOUNDED’ TAWBOOTS <br /> Instructions for Sewing Tucked Frogs and… <br /> This hat measures 5½“ in diameter. Using thread of 1 s c, make a d c. When sewing in the crown, remember to sew between the two shell caps which stay up when hat is tucked in and unused. Next lucky hat 14 d c in 3 distinct spots, gather thread in each d c and sew flat. Next lucky hat, 15 d c in 3 distinct spots. <br /> Next lucky hat 21 d c in same place as
It’s Hang in Wind! Humans cannot comprehend what it does.
Read 18 tweets
Profile picture
Janelle Shane
@JanelleCShane
nothing to worry about. it's only mutant fly strains invented by a neural net
aiweirdness.com/post/186072231… Macho man: likes to punch<br /> Jungle fire: Any of several genes that makes the flies more prone to fires.<br /> Chocolate eye: The fly that drinks and stumbles when eating candy.<br /> Jingle Bells: Does not speak but sings<br /> The last Emperor: Has a huge black mustache<br /> Unicorn faced: A gene that prevents flies from developing fly traits.<br /> Lana: Her eyes are covered by a large purple cloud
the neural net's most intense fly mutants are a bit beyond my artistic abilities but I would love to know what they look like Chocolate Corn Cone: Has six heads. It’s made out of chocolate and corn (in addition to the brains from its victims).<br /> <br /> Lemondogs: The fruit flies from the American Southwest who were responsible for killing the dinosaurs and driving scientists crazy years earlier. As a consequence they evolved into the same kind of insect that scientists now recognize today as the Lommy Beetle, the “mammal-sized flea” from the Southw
the neural net's task was to add to a list of 4 example mutations. This is the 1st task where I've seen a noticeable difference between 117M and 345M GPT-2. 117M doesn't get it. (via minimaxir.com/apps/gpt2-smal…)
The larger 345M often manages. (via talktotransformer.com) Virtus: No. Fishes have no breath.<br /> Tina: There is no food left.<br /> The Dragon Is Burning: Hormones.<br /> The Dragon's Lair: The lair of the dragon.<br /> Prayer of Pain: This is what we will do next.<br /> Prayer of Light: After praying this, I will open fire on anyone.<br /> The Night Star: A star's shadow.<br /> Tina will fly to the sky once<br /> The Night of the Lion: A dream world.<br /> The Night: A fairy's world. <br /> The Night of the Lion: A fairy's loveTiny Hatter: Finesse is too strong.<br /> Grim Reaper: Does not get the joke.<br /> Hawkman: Cannot swim.<br /> Jackpot: Loses an arm and leg in a single punch.<br /> Candy Cane: No longer a criminal.<br /> Raghead: Must die of old age.<br /> Deadbeats: Death must follow.<br /> Shapeshifter: Can turn.<br /> Chameleon: Has become a man of sorts.<br /> The Pigeon: Is not a human.<br /> Wasp Girl: In the jungle, when a man goes missing, flies follow him. No way back.
Read 4 tweets
Profile picture
Janelle Shane
@JanelleCShane
Always ask what training data the AI learned from. Here,
- 16 men and 8 women (probably young & non-disabled)
- pretending to walk with emotion
- auto-transformed to stick figures
- and then 'true' emotions labeled by remote MTurk workers
- ignoring actor's intended emotion
And what about training data for AIs used to identify potential aggression? Or to analyze videos of job candidates?
Every time I see stuff like this I think of CIMON annoying an astronaut with its emotion detection. qz.com/1482839/the-is…
I'm so excited to see this application happen though:
Read 3 tweets

Related threads

Profile picture
Karol Cummins
@karolcummins
💰Banks, Farage, Bannon & Trump

Banks’s affairs have also fallen under scrutiny outside the U.K. Banks, Wigmore, and Farage came to public attention in the U.S. shortly after the 2016 elections.
#Brexit #2016Election
newyorker.com/magazine/2019/…
🔑In London, few days after they were photographed with Trump outside of Trump Tower Banks and Wigmore again met with Yakovenko the Russian Ambassador, detailing their meeting with Trump & passing on contact info for members of Trump’s transition team.
newyorker.com/magazine/2019/…
🔑Yakovenko was expelled from the US as a Soviet spy in 1986.
Read 25 tweets
Profile picture
Duncan Ross
@duncan3ross
So you're a believer in #AI #Ethics? We all are? How come there are so many twitter arguments about it then? The answer (in a Myers/Briggs* style) is because we have different attitudes to certain key questions. Thread. 1/15
Our first disagreement is around People and process (P) vs Technology (T). Do you focus on using frameworks, or on better tech to resolve the issue? 2/15
There are great frameworks out there: @AdaLovelaceInst (turing.ac.uk/research/data-…) and @Floridi, work we've done at @DataKindUK (datakind.org/blog/doing-dat…) with @christinelhenry, the EU (ec.europa.eu/digital-single…) and more 3/15
Read 15 tweets
Profile picture
Audreys Observations
@punkassaudrey
President poised to sign bill creating quantum computing initiative

(When combined with self-learning AI, this technology can be wonderful... or downright terrifying.) engadget.com/2018/12/20/pre…
#AI #QuantumComputing
Think INTEL.

Read 11 tweets
Profile picture
Marie Donelly
@PleasureEthics
It's art up for review in opinions, of course I'm going to gravitate towards something shared with interest, when he didn't say a word for 2 days! Why was he there? #opendata #smartcities #bigdata #ai
I went through every single video he released in hip hop to understand him not saying a word. Easy to work out, he's a very bitter man who has absorbed all his parents immigrant opinions at home #opendata #smartcities #bigdata #ai
His parents are not middle class anywhere & neither is he #opendata #smartcities #bigdata #ai
Read 3214 tweets
Profile picture
inconvergent
@inconvergent
it's 2020. noBlockchain is the new database hype.
it turns out to be csv files, stored directly on the file system.
it's an enormous improvement.
Read 71 tweets

Trending hashtags

#ChestPain #FOSS #government #Rust #robot #JavaScript #MachineLearning #algorithm #postdocalypse #ManOnTheMoon #Sustainability #Disruption #AI #smartcitiea #data #CRISPR #accountability #privacy #GNSS #diagnosis #ClinicalGovernance #FoxP3 #medical #lightspeed #QuantumMaterials
Follow Us on Twitter!

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just three indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3.00/month or $30.00/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!