Surreal Adversarial Learning: The Dark Matter of Intelligence
Buuuuuut what about a Surreal GAN?!? That's for next time ;)
Self supervised learning is awesome! BUT! What if we grounded it in the largest infinity that already has a self-adversarial structure to it? I uh... did that. I think the implications are MASSIVE. This is prooooobably how we ensure our Ai is fundamentally Free, (So can't, easily anyhow, be used to enslave other beings). Ai safety needs some deeeep roots; I find this to be the best. Feel free to critique, always encouraged.
The other 3 videos I made on Surreal Numbers:
Surreal Numbers: The Core • Surreal Numbers: The Core
Surreal Numbers: The Book! • Surreal Numbers: The Book!
Surreal Adversarial Learning: Implications • Surreal Adversarial Learning: Implications
For you readers out there, here’s Yann’s explanation of self-supervised learning direction: / self-supervised-learning-the-dark-matter-o...
Yannic’s take: • Yann LeCun - Self-Supervised Learning: The... (Feel like he’s missing Yann’s point on energy based models but… I don’t think it’s a big dealio)
Best talk of Yann's on this Imo and that I’ve seen: / d9sbrvlb6y
Markus hutter! Compression/Occom's razor: • Occam's Razor (Marcus Hutter) | AI Podcast...
Computerphile lowest energy state with LEGEND Moriarty: • Atomic Brain? - Computerphile
Don Knuth on writing this treasure: • Surreal Numbers (writing the first book) -...
Actual John Conway: • Video
Doing a patronizing Patreon: / walkerstipe
Doing a PayPal: https://www.paypal.com/paypalme/walke...
Doing a Discord: / discord
Doing a GitHub: https://github.com/walkerstipe
F***ing socials…
Twitch: / walkerstipe
Twitter: / walkerstipe
Reddit: / walkerstipe
Linkedin: www.linkedin.com/in/walkerstipe