As the AI scraping war rages on, your personal IP is going to become really valuable. Just like a personal credit score. Your personal browser history—tied to your computer fingerprint—is what is going to give you access to the internet.
With quickly improving diffusion language models and rumors that OpenAI’s new image model is auto-regressive, it will be really funny if we got the optimal models for language and images backwards. Though maybe both are just equally compute and data bound.
When you've done a lot of landscape photography you can spot all of the composites and fakes people try to pass off as real. Few people have the skills to spot them and it's too hard to convince people, so we don't. I wonder what other industries have this problem.
Here's what happened.
1) lost the key to my security locker when I moved
2) learned how to pick locks from a youtube video
3) managed to pick the lock in about 30 minutes
4) decided the locker isn't actually safe and probably shouldn't use it
@ellmetha keep up the good work. Feedback if you're curious, docs look great and are some of the best I've seen. Main thing that stopped me was compilation time. Not sure if there's any way to speed it up :-) Otherwise keep up the great work.