model naming on the lmsys chatbot arena is getting out of hand
RESEARCH
-
Generative AI accelerating metaverse, robotics, synthetic biology
By
–
Discover how #GenerativeAI is set to accelerate #technologies from the #metaverse to #robotics and #synthetic #biology. Explore the #transformative potential of generative #AI in revolutionizing both industry and everyday life.
-
Cat Detection Research Reception in Early Deep Learning Era
By
–
I was a (young) grad student when the Cat Detection work was published and publicized.
I don't think people thought it was a dead-end. But I can explain the general resentment for that result from what was an impressionable few months for me. People didn't really like the -
DistBelief Training System: Clarifying the Cat Detector Paper
By
–
There was a bit of confusing back and forth between myself and @ylecun recently, because in the tweet below, 'it' was referencing DistBelief, but @ylecun thought DistBelief referred to the ICML 2012 "cat detector" paper (
https://
icml.cc/2012/papers/73
.pdf
…), not to the training system called -
Training Extended Context Length for Improved AI Accuracy
By
–
I trained it to extend context length significantly — this has many benefits, including the ability to pull in lots of external information to improve accuracy.
-
HyperWrite Releases Improved LLaMA 3 70B Weekly Iterations
By
–
This iteration of our @HyperWriteAI LLaMA 3 70B is just the start. We will be training and bringing up newer + better models each week. If we do this right, this version will look like an antiquated POS in just a few weeks 🙂
-
Meta’s Sub-$5K LLM Training Outperforms GPT-4 and Opus
By
–
You don’t need billions of $$ to make amazing LLMs anymore. Meta made sure of that. Better than Opus and GPT-4, and we trained it for under $5K.
-
Med-Gemini Achieves State-of-the-Art Medical Imaging Results
By
–
This new paper on Med-Gemini research from @GoogleResearch and @GoogleDeepMind shows how the multimodal and long context capabilities of Gemini 1.5 can be used to get state-of-the-art results on a variety of 2D and 3D medical imaging and genomic risk score tasks.
-
AI Generates Poetry Showcasing Creative Capabilities
By
–
It's not specifically tuned for poetry, but it's still pretty good! In city streets, where crowds of strangers roam,
Each face a world of stories untold;
Their dreams and fears, as real as your own home,
In every glance, a novel left unsold. Their lives, a tapestry of joy and -
Model Training Requires Intensive Weight Adjustment and Data Tuning
By
–
Not great w/o adjusting the weights. This result took weeks of intense data tuning + full-scale training.