DALL•E 2

Some of you will have already seen or read about DALL•E 2 in the past 24 hours, but I can’t resist sharing.

DALL•E 2 is an AI system from Open AI that creates realistic images and art from a text input. The model itself is trained against a vast database of text-image pairs.

Open AI’s CEO, Sam Altman, asked his Twitter followers to submit strings of text and he’d reply with the image generated by DALL•E. The results are nothing short of astounding.

Text: A rabbit detective sitting on a park bench and reading a newspaper in a victorian setting – source

Text: A painting inspired in Banksy’s art showing a human-machine interaction – source

In fact, the more words, the better!

Text: A shipping container with solar panels on top and a propeller on one end that can drive through the ocean by itself. The self-driving shipping container is driving under the Golden Gate Bridge during a beautiful sunset with dolphins jumping all around it – source

I also found Sam’s reflections interesting to read.

He acknowledges the possibilities: this is perhaps the beginning of a new computer interface trend (i.e., you say something to a machine in natural language, and it does it).

He recognises the danger: the model is already powerful enough to imagine the downsides such as disinformation.

He also reminds us that AI predictions are so difficult to make: 10 years ago we believed AI would replace physical labour, cognitive labour, and only then tackle creative tasks. It seems now that the reverse could be true.