If you can imagine and describe it, DALL-E 2 can create it. Meet the AI tool that’s about to revolutionize the world of creative images.
A bowl of soup… that’s a portal to another dimension… in the style of Basquiat. It’s not real. Nor has it ever been depicted before by a living artist. Yet as quickly as you can type the description, DALL-E 2 can create it.
Developed by OpenAI, DALL-E 2 has been programmed to create images based on simple text descriptions. All you have to do is describe the scene, context, and any other details you like (from camera angle to the artistic style) and it will generate multiple images to choose from. The results are impressive, sometimes hilarious, and have led to a number of creations that have gone viral on Twitter.
The technology is based on GPT3, and built upon an enormous collection of artwork spanning 40,000 years of our civilization. At first, it might be easy to miss just how big an impact a technology like this could have on our society. After all, can’t you just do a Google image search of almost anything these days and find results roughly in line with what you typed?
Except, DALL-E 2 isn’t based on any predefined data. It’s not simply finding images that already exist. It’s not even about mixing together old images that have been created in the past. No, DALL-E 2 is able to create completely original pictures that have never existed before. It doesn’t take things from the past, it learns from them. Like a real artist. Not only that, it’s also learning to predict human judgement, so it knows what is most likely to please the human eye.
If DALL-E 2 is able to create something original that humans find beautiful, one might ask whether there’s a chance that it could replace artists altogether. We’ve been hearing about robots taking our jobs for some time now. In February 2021, McKinsey Global Institute predicted that 45 million Americans—one-quarter of the workforce—would lose their jobs to automation by 2030. But when we talk about automation, we imagine self-driving taxis and delivery drones. Not art.
There is also another more sinister side to the technology that requires attention. A photo realistic image is funny when it’s teddy bears dancing on the moon, but what if it shows a real person saying something they never said, or doing something they never did?
When it comes to ‘deep fakes’, as AI technology continues to advance, so do the dangers. The possibilities for more sinister uses for this technology has led to questions being raised on how this tool can be released safely. DALL-E 2 is well aware of this and acknowledges that without sufficient guardrails, the technology could be used to ‘generate a wide range of deceptive and otherwise harmful content’. To mitigate these risks, they have implemented a safety policy that includes advanced input filters, limited access, and various monitoring and reporting mechanisms.
We can’t know for sure what the future holds, but one thing is for certain, DALL-E 2 (which is not yet available to the general public) has the potential to make some major waves in the creative industries. Is it purely a threat? Or would it be better to acknowledge it as one more in a long line of artistic tools – including pens, computers and Photoshop – that won’t eradicate our creativity, but amplify it?