Introducing: How to Generate (Almost) Anything
Can AI inspire us to push the boundaries of creativity?
This article originally appeared at How to Generate (Almost) Anything.
In the midst of ‘AI awakening’ where machines are becoming good at many ‘human’ jobs, people are worried that AI will ultimately lead to mass unemployment by replacing them. On the contrary, we believe that we can achieve the most creative and productive outcomes when humans and machines work together to enhance each other’s complementary strengths and skills.
By augmenting human capabilities and pushing the boundaries of creativity, can AI inspire us to create things that wouldn’t have existed otherwise? A dress designed with a crazy hat, a pizza made with shrimp & jam or a scent that has never been smelled before? We are a bunch of scientists and artists aiming to test the limits of human-AI collaboration to generate (almost) anything the human mind can (and can’t) imagine. Inspired by MIT’s most celebrated How to Make (Almost) Anything class, we introduce How to Generate (Almost) Anything!
Every week, we generate something with AI and present an exemplary creation by collaborating with an artist, artisan or a scientist to bring some of the AI’s dreamed-up creations into life.
But also we want you to help us! Try out AI’s designs from the project pages by adding your own perspective and insight to create something even better. Take a picture or a video of your creation and share the results with us. We will add them to our website for everyone to see!
Also don’t forget to subscribe to our Youtube channel, follow us on Twitter, Instagram, or sign up our newsletter to get updates!
Frequently Asked Questions
- Where does the inspiration of the project come from?
How to Make (Almost) Anything is one of the most popular classes at MIT, taught by Neil Gershenfeld (the father of the maker movement). It is also one of the toughest classes to get in, and there is a long waiting list every year. Most of us were among the lucky students who were able to take the class last year where we learned a new technology every week (everything from laser cutters to 3-D printing), and made something with it. In the midst of AI awakening, we wanted to create something similar where we generate something every week!
2. What are the goals of this project?
- Human-AI collaboration: Every week, we train an AI to generate something, and invite an artist or an artisan to collaborate with the AI. In an era where people are worried that AI will take over their jobs, we want to show to the public that they don’t need to fear from AI, but use it as a tool to complement their strengths and skills to create the most creative and productive outcomes.
- Pushing the boundaries of creativity: We believe that AI can augment human capabilities and inspire us to create things that wouldn’t have existed otherwise. In the experiments we made so far, we designed a prom dress from the future, cooked a pizza with shrimp & jam, created a scent that has never been smelled before, and painted the walls of Graffiti Alley in Cambridge — all inspired from AI-generated content!
- Democratizing AI for non-practitioners: So far, AI-driven creativity has been within the reach of AI researchers or skilled programmers. We want to democratize the usage of AI for creativity. Every week, we show an exemplary human-AI collaborated work on our Youtube channel and we share hundreds of AI-generated designs on our project pages for the public to try them out. This is a preliminary step we take towards democratizing AI for non-practitioners, but we are currently working on solutions to allow people play with cutting-edge deep learning models and help them generate (almost) anything!
- Bringing AI’s hallucinations into reality: Generating stuff with AI is not new: researchers and programmers have been working on creative applications of AI for a while. But usually most of the creations AI dreams up don’t come into reality. In this project, we want to make it real: how does that pizza recipe actually taste like? How does an AI-generated perfume smell like in real life? Or how that dress AI-dreamed up would look like on a real person? With this in mind, so far we bring AI-dreamed up music, pizzas, fashion, perfume, graffiti into reality. Check out the projects below, and take a look at our Youtube channel if you want to see us in action!
3. Is this a project funded by MIT?
This is an initiative entirely run by students, and currently not funded by anyone or any organization. But feel free to drop us a line if you want to sponsor us (more GPUs are always welcome).
4. How do you name the AIs?
Most of us are science fiction fans (and amateur fiction writers). We name the AIs we trained with relevant characters from sci-fi books. So far, our AIs are:
- Halleck in Episode 1 (AI-generated music), named after Gurney Halleck, a musician from the Dune series.
- Strono in Episode 2 (AI-generated pizza), named after Strono Tuggs, a cook from Star Wars.
- Garak in Episode 3 (AI-generated fashion), named after Elim Garak, a skilled tailor from Star Trek.
- Crompton in Episode 4 (AI-generated perfume), named after Alastair Crompton, a skilled nosologist from Crompton Divided.
- Chopper in Episode 5 (AI-generated graffiti), named after Chopper, a graffiti artist from Judge Dredd.
- Wintermute, the spokesperson for all the AIs we collaborate in our projects. Similar to his namesake from Neuromancer, its goal is to become a super-intelligence but currently playing along with us.
Episode 1: Human-AI Collaborated Music
In the honor of the father of artificial intelligence and MIT Media Lab’s founding member Marvin Minsky’s birthday, we pre-launch our series with a duet between Marvin and an AI (trained on thousands of arcade songs from 80s and 90s). Then, we played the Marvin-AI collaborated song on the grand piano in MIT Media Lab’s basement floor. Visit the project page, or watch the Youtube video below:
Read more about this project here.
Episode 2: Human-AI Collaborated Pizza
We trained an AI on hundreds of artisan pizza recipes, and made it to generate dozens of new (& weird) pizza recipes. Then, we collaborated with Tony, the chef and owner of Crush Pizza, one of the best artisan pizza places in Boston, to cook four of them on wood-fired brick ovens. Yummy! Visit the project page, or watch the Youtube video below (watching with closed captions on is highly recommended):
Read more about this project here.
Episode 3: Human-AI Collaborated Fashion
We trained an AI on thousands of dress designs from vintage sewing pattern magazines, and made it to generate new ones. Then, we re-imagined how the dresses would look like in real life, and the results are fascinating! Visit the project page, or watch the Youtube video below (watching with closed captions on is highly recommended):
Read more about this project here.
Episode 3: Human-AI Collaborated Perfume
We trained an AI on thousands of perfume notes, and made it to generate new ones. Then, we made two of them in the lab to bring AI’s dreamed-up scents into life! Visit the project page, or watch the Youtube video below:
Read more about this project here.
Episode 5: Human-AI Collaborated Graffiti!
We trained an AI on thousands of pictures of graffitis taken all over the world, and made it to generate new images.Then, we collaborated with graffiti artists to paint some of them in the famous Graffiti Alley, Cambridge. Visit the project page, or watch the Youtube video below:
Read more about this project here.
Episode 6: Human-AI Collaborated Chocolate Truffles
We trained an AI on hundreds of chocolate truffle recipes, and generated new ones. Then, we collaborated with MIT Lab for Chocolate Science to make four recipes and tested the truffles in public! Visit the project page, or watch the Youtube video below:
Read more about this project here.
Next week: Human-AI Collaborated Viruses!
Next week, we are going to generate protein structures of viruses. Stay tuned!
Subscribe to our Youtube channel, follow us on Twitter, Instagram, or sign up our newsletter to get updates!