TwoCents | I tried Dall.E last year, here's what I think.

The image above was generated with DALL.E - I typed in "A blue sports car in a small village" and that's one of the images I got. The other variations it generated were pretty cool too. So by now most of you know what DALL.E is and have probably tinkered around with it - but for those of you who don't, it's an AI tool created by OpenAI that let's you enter short descriptions in a prompt then generates several images relating to your description. You get 50 free credits and for every set of images it generates for you, you use 1 credit. I spent some time with it and the results were pretty interesting! Infact there's a post on TwoCents whose image was created by DALL.E, here's the link to the post https://twocents.space/insight/naira-redesign-effect-on-the-economy-and-people-it-is-no-longer-news-that-some-notes-of-the-nai-450/. As at the time of writing this I have about 20 credits left, so yeah, I messed around with it a bit. Needless to say I think the guys at OpenAI have done an amazing job and massive credit to their ML team. I'm not an AI/ML expert/researcher so I can't really speak to some of the technical aspects of the model they're using, but I think they've created something quite phenomenal, even though I think it still needs a lot of improvements before it becomes a serious tool for creators and creatives. Like I said above, one of the posts on TwoCents has an image that DALL.E created - for that one I typed in "New Naira notes" and that's one of the images I got; thinking about it now, I should have added more specificity like "new Nigerian Naira notes" and I'd probably have gotten an even better set of images, even though I'm pretty satisfied with the one in question. DALL.E also allows you to upload an image and it spits out variations of the image - I wasn't quite impressed with some of the variations it gave me when I uploaded pictures of people - the variations looked really weird and in some cases, there was almost no resemblance between the original image and the variations. There was this particular image of Peter Obi I uploaded and what is spat out made me cringe - another one was an image of some protesters from the 20.10.20 Lagos protest and the variations looked like inbred ogres from Apokolips (Darkseid's planet from the DC Universe). Not all the images with people had bad variations, there were a couple with female faces that had pretty impressive results. The variation results were more impressive when the image was something like a bar of chocolate, a map of the world or a handbag. I also like the fact that they've put in a lot of measures to ensure you can't generate explicit images or anything pornographic. Overall I think DALL.E is an amazing tool and it's only going to get better. The biggest thing for me with DALL.E and ChatGPT is how long they had worked on this for - OpenAI has been around for a really long time, many people might not know this but they have, and Elon Musk was heavily involved in it's leadership but stepped aside at some point. Speaking of ChatGPT, they recently released a pro version so that should be quite interesting for people who already use it heavily. Earlier in this post I said that one post on TwoCents has an image generated with DALL.E, well, there's actually 3! Here are the links to the other two along with the description I entered: "The Spitting People Of Massai Tribe" - https://twocents.space/insight/the-spitting-people-of-massai-tribe-456/ "Futuristic neon lit African city at night" - https://twocents.space/insight/the-real-wakanda-is-the-democractic-republic-of-congo-sometimes-life-imitates-art-other-times-435/ Let me know what you think of DALL.E in the comments if you have tried it. Thanks for reading!

More insights from Cyrus Majebi