100% agree, "the proof is in the pudding". It has to actually work. I will say that I think the technology exists today to actually make it work at the needed threshold. Actually making it work is still difficult. But 6 years ago I would have said the technology does not exist.
RESEARCH
-
Smartphones, Children’s Brains, and Neural Implants: Neuroscience Report
By
–
Hyper intéressant, le rapport de l'utilisation des smartphones par les enfants, et la neuroscience avec les implants.
-
How AI is Shaping the Future of Education
By
–
Check out my @forbes article for insights on how AI is shaping the future of education https://
forbes.com/sites/ronschme
lzer/2024/05/28/how-ai-is-shaping-the-future-of-education/?sh=1346ac0bc9ec&utm_source=dlvr.it&utm_medium=twitter
… -
Plateaus, Not Minima, Are Gradient Descent’s Real Challenge
By
–
The real problem in gradient descent isn’t local minima, because they’re extremely rare. It’s not saddle points either, because any small amount of noise gets you out of them. The real problem is plateaus.
-
Evolution Created Us to Accelerate Itself, Then We Created AI
By
–
Evolution created us to speed itself up, and then we did the same with AI.
-
AI and Climate: Prioritizing Real Issues Over Distractions
By
–
It distracts from the real issues we have to solve, both AI related and climate related.
-
LoRA Training 4-8x Faster Than Full Fine-Tuning
By
–
4/ Faster to train! Training with LoRA can easily be 4-8x faster than full fine-tuning.
-
How LoRA is Transforming LLM Fine-tuning Techniques
By
–
1/ Curious how #LoRA is changing fine-tuning? Read our latest blog post to learn more! #llm #finetune
-
LoRA Achieves Comparable Performance to Full Fine-Tuning
By
–
2/ No trade-off in performance. A 2021 study demonstrated comparable performance between LoRA and full fine-tuning.
-
LoRA Training Parameters: Minimal Memory Footprint at 0.2%
By
–
3/ Extremely low memory footprint. LoRA trainable parameters are only 0.2% of total amount of model parameters.