If you ever wanted to see how good (or not) AI and neural networks are at accurately identifying things, you should check out a new iOS app called AI Scry. The app was developed by an art and technology studio named disk cactus (their real name is just an emoji of a floppy disk and a cactus plant, which won’t show up here) who wanted to create an easy way for people to understand the problems and benefits of machine learning.
The problems with machine learning are more obvious with AI Scry than the benefits. Point your iPhone at everyday objects and the app will read back what it thinks it’s looking at. AI Scry sometimes gets it right, but most of the time comes back with nonsense—although it’s adorable nonsense.
“It can tell the difference between the contents of a fridge and a park full of trees, for example, or spot that this is a car and that is a person. And occasionally, it makes imaginative leaps that seem ingenious,” notes The Verge. “At one point the app described a particularly dense and bushy-looking bush as ‘a close-up of a garden of broccoli,’ which borders on the poetic.”
The neural network powering AI Scry is the open-sourced Neural Talk, developed by Andrej Karpathy, a computer scientist from Stanford University. And though the app and its neural powered network is far from perfect, Sam Kronick, one of the app’s developers, tells The Verge that it’s a good way to introduce people to the world of machine learning without being “too robotic.”
Kronick says that it’s important for people to understand just what machine learning is capable of, considering it is increasingly used to compute credit scores, and do less critical things, such as creating Spotify playlists.
“When you set one free and unconstrained in the world (as with AI Scry), it’s plain to see that it has a ‘mind’ of its own,” Kronick told The Verge. “[It’s] complete with weird quirks and idiosyncrasies that can be traced back to the chain of humans who designed, programmed, and trained the system.”