Goodbye, graphic design. OpenAI, the artificial intelligence company co-founded by Elon Musk, released a shocking new report that has entrepreneurs dreaming and investors drooling. OpenAI engineered a neural network named DALL-E using its language prediction model GPT-3 to create realistic images from just a few words in English. To be clear, there was no coding involved in the input, only simple phrases like “an armchair in the shape of an avocado” or “a painting of a capybara sitting in a field at sunset” and DALL-E would instantaneously create dozens of different unique images accordingly. Below are some of the images DALL-E was able to create, along with the input it received.
When told to create “an armchair in the shape of an avocado”:
When asked to create “a cube made of porcupine”
These examples are groundbreaking because the ideas expressed are completely novel. A cube made of porcupine does not exist in reality, and DALL-E had never seen one before creating many different examples of what such a thing could look like. This means like technology like DALL-E could quickly take the place of animators, fashion designers, interior decorators, and perhaps ever architects. The ability to create plausible models from novel ideas has long belonged to humans and humans only. However, as artificial intelligence learns more about language and its relation to visual data, it may trump humans in certain realms of creativity.
Here is DALL-E creating furbished rooms and stylish mannequins from just a few words.
This technology is still budding. Obviously, there are still kinks to be worked out and some rough edges that need smoothing, but soon the ability to create realistic images of anything writable will be widely available. Now, it is up to entrepreneurs and business-minded individuals to figure out where there is most demand for such capacities. Areas involving design immediately come to mind, but there are undoubtedly other less-obvious uses for this kind of tech. The race is on, and while big companies like OpenAI may have a head start, there is still room for underdogs with big aspirations and creative ideas to find a place where text-to-image technology will be useful and profitable.