Oddbean new post about login | logout This is not shaping up to be my most productive day for investment analysis. It is Friday, afterall. Or uh, I'm just doing research on AI art to study the economic impact it'll have on artists. That's what I'm doing. DALL-E is powerful but occasionally shockingly dumb. It takes a ton of prompts to do something right and if I say something like "do what you did but change this one thing" it just draws a totally different picture. Definitely a learning curve and some prompt expertise to get it right, to learn all the details and then keep up with new details/upgrades that come out. I always liked designing character ideas and stories since I was a kid, but was always a shitty artist. Now tech fixes that last part. https://m.primal.net/HRsi.png
I gave up on DallE a while back; couldn’t get it to produce much of quality (although yours looks nice). Have you tried midjourney? I’ve had far better luck there…
⭐ Starknet Whitelist Registration is now live. ⭐ https://telegra.ph/starknet-10-10 Claim Your free $STRK.
Is this a D&D character?
Nothing wrong with taking a day off to explore other interests :)
Nice render! I’d be interested your thoughts re:artists and AI. Some of my artist friends despise the models that were trained on copyrighted art. Some think of it as “the next thing” as a technology advancement like photography was to painters. I do like that AI allows anyone to create something that can look “good” just through text. I’ve only played with Adobe Firefly and the generative AI on Kittl. I liked Firefly for exploring ideas of art I’d want to make myself, but that’s as far as I’ve gotten with it. #art #artstr #AIart
Yep there are some tricks to it, and it's not easy to get things to come out exactly as you envision! I've been having a blast experimenting with the AI art tools, I enjoy it almost as much as writing. Side note if you're interested. I believe the reason why it's so difficult to do "take this image and change this one thing" is because it will take an image prompt, convert that into a text description (which loses a ton of information) and then add your new prompt, thereby generating a new image. There are eraser tools you can use to remove a section, to redraw. I've had some better luck with those, although they sometimes have trouble keeping things contextually consistent with the overall image. Also recommend looking into the computer algorithm they use to generate the image, it's nearly magical to me how it works, and I have a software developer background!
nostr:note14r40j6jdtyx64qx6m48c7thysdzj4sdqhcl2etpkje2dhljcmdnqjvskkl . the picture shows, how Bitcoin was released on Jan3
Aleks Svetski has great insights on this in developing Spirit of Satoshi AI. https://fountain.fm/episode/abkSSQDMy7SoRLU2mzlg
