Many GenAI models are based on Markov processes. By parameterizing them w/"generators," they can:
1. Unify & universally characterize Markov models
2. Train them at scale
3. Combine them to build multimodal models or superpositions
4. Train new models, e.g. jump models
RESEARCH
-
Markov Processes Parameterization Unifies Multimodal GenAI Models
By
–
-
FrameJump: Novel Jump Models for Protein Generation
By
–
Their experiments show:
— MIT CSAIL (@MIT_CSAIL) 8 novembre 2024
1. Jump models are an unexplored model class w/promising results.
2. Combining models via Markov superpositions improves generation results. Check out FrameJump, a novel protein generation method — they can make amino acids jump to a protein! pic.twitter.com/vUNm2XxHs6Their experiments show:
1. Jump models are an unexplored model class w/promising results.
2. Combining models via Markov superpositions improves generation results. Check out FrameJump, a novel protein generation method — they can make amino acids jump to a protein! -
Unified Theory of Generative AI: Generator Matching Framework
By
–
Can we design a generative AI algorithm that works for any data type? Together with a team at FAIR, CSAIL researchers propose a "unified theory of GenAI" called "Generator Matching" (GM). GM brings together a range of GenAI models, allowing researchers to build models for any
-
Team of Researchers Reviews Feedback to Improve Developer Models
By
–
We have an amazing team of researchers and folks like myself who pour over these responses to make our next models even better for devs – trust me, we read all of it!
-
Small Models Performance for Spam Classification Task
By
–
Sebastian Raschka
How about these wrt to small models on a specific task (here: spam classification) -
Rethinking AI Implementation Through Collective Intelligence Lens
By
–
Are we thinking about AI implementation all wrong Professor @criedl
's fascinating article in @HarvardBiz challenges the conventional automation vs. augmentation debate and presents a revolutionary perspective: focusing on #CollectiveIntelligence. Here are the -
SG-I2V: Self-Guided Trajectory Control in Image-to-Video
By
–
SG-I2V
— AK (@_akhaliq) 8 novembre 2024
Self-Guided Trajectory Control in Image-to-Video Generation pic.twitter.com/fCnHHUsRWrSG-I2V Self-Guided Trajectory Control in Image-to-Video Generation
-
DynaMem: Dynamic Spatio-Semantic Memory for Mobile Manipulation
By
–
DynaMem
— AK (@_akhaliq) 8 novembre 2024
Online Dynamic Spatio-Semantic Memory for Open World Mobile Manipulation pic.twitter.com/ZOJ0hI7n5kDynaMem Online Dynamic Spatio-Semantic Memory for Open World Mobile Manipulation
-
DimensionX: Generate 3D and 4D Scenes from Single Images
By
–
DimensionX
— AK (@_akhaliq) 8 novembre 2024
Create Any 3D and 4D Scenes from a Single Image with Controllable Video Diffusion pic.twitter.com/QCYIyMbbrFDimensionX Create Any 3D and 4D Scenes from a Single Image with Controllable Diffusion
-
OpenCoder: Open Cookbook for Top-Tier Code Language Models
By
–
OpenCoder The Open Cookbook for Top-Tier Code Large Language Models