Naturally one ought to be curious
By James Vincent Jul 10, 2019, 8:00am EDT
(...)
Runway began as Valenzuela’s thesis project at the Tisch School of the Arts at New York University. After getting enthusiastic feedback from the AI art community, he decided to take the program mainstream, asking two school friends to come on board as co-founders and gathering seed money from NYC and Silicon Valley backers. The company was incorporated last December with a beta launch following this January.
Valenzuela straddles the fields of art and code and says he wants to bridge these two worlds, empowering non-coders to use machine learning models and, in turn, connecting researchers to the people who will benefit directly from their work.
In a blog post Valenzuela wrote last May, he compares the current AI art scene to painting in the 16th and 17th centuries. At that time, the act of simply storing and using paint was something of a craft secret, with painters relying on esoteric techniques involving pigs’ bladders and string. But with the invention of the paint tube in 1841, the craft became more accessible. It was also easier to conduct outdoors, leading to new styles and movements. (...)
The company is not the first to make AI models easier to use of course. But earlier examples — like Lobe, which let users train AI systems using a visual interface before it was bought by Microsoft — have focused on business use cases, rather than creative ones.
Runway’s target market is obvious when you load up its store front, which lets you browse a range of models that run the gamut from text generation to motion tracking.
You click to see the details of each model, click to add it to your workspace, set up your inputs and outputs, then start the system running. There are hooks to connect these outputs to other apps (so you can send ML-processed images to Photoshop, for example), and users can import new models directly from GitHub with just a few lines of code....
