Post 2: Dall-E đ¨
Published:
Â
Recently, an artificial intelligence (AI) lab called OpenAI (founded by Elon Musk) announced the update of one of its most powerful AIs called Dall-E v2 (i.e. Salvador DalĂ + Wall-E).
Dall-E takes any idea you give it in text form and turns it into an image. It can even mimic the style of various famous artists. (Here and in the comments, Iâll share several of the images Iâve come across on Twitter and liked, as well as a gallerycreated by Dall-E of +1000 images of robots and plants).
EThese types of AIs are not new; however, Dall-E is unique in its âcreativityâ power. Although the paper details some of how it was trained, we donât really know many relevant details. But we can say that these types of AIs are trained using a combination of two models, one that understands text and another that learns to translate it into images. If you want to learn more about how it works, I highly recommend the following video: ⌠And if you want even more, this list of videos from the same channel.
Dall-E is not available for public use and it seems it will stay that way, but you can request to use it for academic or outreach purposes (for now). Its implications will be enormous, as in about 5 years or less, we will probably see these types of technologies integrated into our phones and software. It could be a threat to those working in related fields, such as graphic design. And it also reignites the debate regarding the wave of misinformation we generate from false posts that we share without verifying what is said or the origin of the photos, which can now even be modified simply with Dall-E.
Here more images:
Someone even went through the trouble of creating a mini digital gallery of several images created by the Dall-E v2, and is so cute.