AI: Papers
The Instant ID mentioned in Propaganda section.
Testing LLMs
Quantum denoising in diffusion
Enhancing Cancer detection with AI
A method to save ChatGPT from jailbreak attacks
Fixed Point Diffusion Models
Pixart-a is a new image creator
Language Models for Quantum simulation
Artificial Vector Retrieval
Testing LLM generation
Mobile Diffusion is designed to work on a phone.
AI teaching AI
Generative AI will save jobs and make everyone fast says Stanford Study
UniSim simulates the full real world
Enhancing LLM with controls
12 pinch points for RAGs and a few solutions to fix them
AI: 3D
DJI releases it's own 3D modeling program
ProteusNeRF is 3D aware
Triplane Gaussian fields
Pretraining cameras for NeRF
Dreamspace is an interior creator
Diffusion Signed Distance Function is a really boring but important model making
Omniverse AECO Demo Pack for digital twins beginners
Dreamcraft 3D is a 2D to 3D program
ZipNeRF uses a grid based system
Human motion control in AI
3D scanned African Holy Ground
Instamat has new procedurally created materials
_3D: 3D GPT
3D-GPT is what it sounds like.
It's useable through Blender
Another article about 3D GPT
A paper on the subject
How many of these do I have?
_3D: Funny Names
GARfield is a depth hierarchal grouping system in 3D scans. You click on a typewriter, and everything for the typewriter is grouped together.
SMERF by Google has brought NeRF back
https://youtu.be/Olos5h8fmXE?si=JDSql41IhIh-z5Vk
:: Bytedance
Bytedance has created a really amazing depth analysis program. Here is an overview on LinkedIn
Here is the page talking about it
AI: 4D
Another 4D creator
4D Gaussian Splatting
The paper on that
Turning movie scenes into 3D moments using AI
AI: Video
Show-1 combines pixel and latent diffusion for video
DragNUWA uses several factors to make videos
A cost analysis of making a video using AI vs the regular way.
Short film competition
AI: Sound
Voice cloning is now being publicly investigated.
https://www.instagram.com/reel/C25iU1zu2Is/?igsh=MThydmVmbnhsaWZvaQ==
These lists are good. It looks like a lot of this was being worked on underground but what caused all this A.I tech to spring out into the wild does anyone know?
Machine Learning has been around for a while. Training the machine is about 8 years old. Having the machine train itself is over 20. NVidia has been showing off stuff for over a decade.
Schools have had it, but simply dabbled till someone got enough money to bring it to the public via NVidia cards. It also helps that the Bitcoin rage died off and all these cards were available.
I think LLMs with tons of extra training really helped it a long the way. Large Language Models are just databases with knowledge on how the words work. So it finds a pattern, and works with that. You need a new poem by Wadsworth? The LLM better have his old poems to work with.
The ability to train the data ourselves is also really useful. Loras and other training only needs a few pictures, and a label. The database searches for others in similar styles and copies them.
So we had all this, and things being labeled for years. The AI community knew how it worked, but didn't jump on it till one got popularity outside of their community