1/10 📢 With recent calls for a "Manhattan Project" for AI safety, it's crucial to remember: the Manhattan Project involved 130,000 people.
Yet, only a few hundred people globally are currently working DIRECTLY on AI safety. This needs to change. 🧵
Enter in a GitHub URL for an old bioinformatics repo, and instantly get a working coding environment you can interact with in your browser.
If you want to join the beta, fill out this Google Form: https://t.co/HwT5qMi5je
A lot of people have been saying that the messed-up hands in AI Art can be easily solved with negative prompting…
…but it turns out all we needed was reverse psychology?
The Stable Diffusion lawsuit seems to fundamentally misunderstand the swiss roll dataset as being an image of a spiral, rather than a distribution of 2D datapoints. Pretty embarrassing mistake.
So does this mean the artists whose work is safest from being copied by generative AI…
…are simply those whose work is so revolting that their name gets banned from being used in a prompt?
going to the barbershop after doing AI art prompts for twelve hours straight and when they ask how I want my beard I say “handsome young bridegroom, baltic wedding, neat, masculine, highly realistic 8K unreal engine”
Well done on DALL·E mini, @borisdayma and team!
https://t.co/bF3qkZGawg
(by the way, this is the best result I've seen so far from ANY image-generation model in rendering a Soviet ekranoplan)