So I spent my Sunday evening training a neural network to generate 🎥 movie plots, and the results are...
Intriguing and hilarious.
Read on for examples.
1/ Movie plot involving 👽 ALIENS:
"Aliens land in the California coast to find a way to fight against the machines."
"Aliens land on Earth, kidnap young people, make some money selling drugs and end up in the desert where they live."
2/ "Aliens land in our city and kidnap people to keep them in it. Their leader is an alcoholic cop, and the people of the city try to get him into custody and take revenge."
"Aliens land on earth and the moon in pursuit of an extraterrestrial scientist"
3/ Plots involving 🤖 ROBOTS
"Robots from all walks of life are terrorizing Europe."
"Robots are on a mission to steal a precious chest to retrieve an atomic bomb in the midst of a global war."
4/ "Robots and an army of aliens fight off a band of renegade spies in an effort to find a legendary relic."
"Robots from Earth go on a mission to Mars."
5/ "Robots in a small French town run into a mysterious terrorist organization"
"Robots go on a quest of discovering a secret weapon which is able to use people to transform them into giant dragons."
6/ Plots involving 👮 TWO COPS
"Two cops try to help a blind man with a difficult case, but when the case is discovered, the blind man is kidnapped and held by a gang of armed men in order to be reunited with their loved ones."
7/ "Two cops, who are on trial for their crimes, fall in love despite their differences."
"Two cops try to solve an identity crisis in the town of Otello."
8/ Plots with 👨🔬 scientists
"Scientists discover that the bodies in a car are very radioactive and a series of mysterious events threatens the balance of the universe. In an attempt to prevent the nuclear apocalypse, a new government sends a team of spies to uncover the source."
Want movie plots to be generated?
Send me a beginning of the plot (e.g. "Scientists discover" or "Aliens") and I'll generate one for you.
This is not GPT-3, by the way.
Finetuned GPT-2 on IMDB movie dataset.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
Turing’s Dream first batch - who is in it and what they’re upto.
🧵
1/ Praveen Chavali - @praveen_chavali is exploring the math of neural networks, and is trying to build a black box optimization method for compressing large models into smaller models.
At yesterday’s tech deep dive, he showed why GANs never converge.
@praveen_chavali 2/ Mehul Goyal - @observerforever is a former hedge fund guy who is now exploring how to model time series data using deep networks.
Yesterday, he explained the problem formulation of predicting sharpe ratio via a reinforcement learning kind of a setup.
1/ I love thinking about thinking. Give me a research paper on rationality, cognitive biases or mental models, and I’ll gobble it up.
Given the amount of knowledge I’ve ingested on these topics, I had always assumed that I’m a clear thinker.
2/ Recently, though, it hit me like a lightning strike that this belief is counter-productive.
That’s because is you “know” that you’re a clear thinker, you’re less likely to suspect that you might be missing something big in your thought process.
Effective technique for not getting involved with thoughts and emotions.
Correction: it’s acceptance commitment therapy.
By the way, not fusing with thoughts and emotions is the core of mindfulness and is pretty powerful.
Thoughts and emotions that bubble up to your consciousness is mostly clickbait - they got selected precisely because they’re exaggeration’s fabricated to make you pay attention.