Well, technically, my first paper on Joint Embedding Architectures (AKA Siamese nets) is from NIPS 1994.
but that paper and many subsequent works use sample-contrastive learning and no predictor.
The JEPA idea trained non sample-contrastive losses (Barlow Twins, VICReg, MCR2) or
INNOVATION
-
Joint Embedding Architectures and Non-Contrastive Learning Methods
By
–
-

Waymo Self-Driving Cars Outperform Humans on Road Safety
By
–
So much for the idea that self-driving cars are hitting walls.
They are not hitting obstacles nearly as much as human drivers. Yes, Waymo cars only drive in some areas that are fully mapped, they use all kinds of sensors, and they have required a decade of data collection, -
Elon Musk Unveils Grok AI Bot With Sarcasm and X Data Access
By
–
Elon Musk unveils new #AI bot ‘Grok,’ says it will have sarcasm and access to X information https://
wsj.com/tech/ai/elon-m
usk-says-his-new-ai-bot-grok-will-have-sarcasm-and-access-to-x-information-b4e169de?st=3xkdzh48i8h9zi9
… via @WSJ -
01.AI Launches Yi-34B Open-Source LLM Model
By
–
http://
01.AI is a bold but long-awaited endeavor for my pursuit of AI over 4 decades. Proud to introduce the world's top open-source model Yi-34B as our first release to the developers' community encouraging fantastic LLM projects with a moderate-size, high-performing -

Waymo Self-Driving Cars Demonstrate Superior Safety Record Compared to Humans
By
–
Waymo self-driving cars already appear to be much safer than human drivers: 1/4 the accident rate of the average person insured by SwissRe. And as @SebastianThrun reminded me last weekend when we shared a Waymo ride in San Francisco, unlike humans, self-driving cars get better
-
Perplexity Open to AI Collaboration with Key Figures
By
–
Sure, we’re very much open to bringing it up on perplexity if @elonmusk and @ibab_ml are.
-
OpenAI DevDay Features Questioned Against Startup Innovations
By
–
"OpenAI DevDay could be the biggest AI event of 2023." Hmm. Why do you think so? None of them are new. They are just some features other startups are doing.
-
Gaussian Splatting: Faster 3D Rendering Than NeRFs
By
–
Remember NeRFs? Gaussian Splatting is faster, editable, and readily integratable. We're talking 100 fps rendering performance on a modern GPU 🤯
— Bilawal Sidhu (@bilawalsidhu) 5 novembre 2023
It's a big step towards democratizing photorealistic 3D content creation.
Here's the full 8 min deep dive: https://t.co/g7hWm4R7qc pic.twitter.com/pK9W9jnAHWRemember NeRFs? Gaussian Splatting is faster, editable, and readily integratable. We're talking 100 fps rendering performance on a modern GPU It's a big step towards democratizing photorealistic 3D content creation. Here's the full 8 min deep dive: https://
youtu.be/sQcrZHvrEnU -

DeepSpeed-FastGen: 2.3x Throughput Improvement for LLM Serving
By
–
Introducing DeepSpeed-FastGen V/ @MSFTDeepSpeed **************
Serve LLMs and generative AI models with
– 2.3x higher throughput
– 2x lower average latency – 4x lower tail latency
w. Dynamic SplitFuse batching Auto TP, load balancing w. perfect linear scaling, plus -
Scale AI Offers RLHF at Scale for Language Models
By
–
if only there was a way to get RLHF at scale for your language models, trusted by all of the leading LLMs (seriously, ping us @scale_AI if you want to build durable differentiation for your LLMs via RLHF…)