Congratulations Gecko team. Evaluation is one of the most important aspects of Gen AI.
RESEARCH
-
LeRobot: Hugging Face launches first robotics library
By
–
Welcome LeRobot! the first robotics library at Hugging Face Over the past months, we saw impressive research breakthroughs in robotics (ALOHA, diffusion policies, UMI and so many) enabling robots behaviors previously thought impossible to train with limited quantity of data
-
AI Challenges Us to Become More Human
By
–
Why AI Challenges Us To Become More Human Learn why embracing #AI could be the key to unlocking human potential and fostering a future where #technology enhances, not replaces, the essence of #humanity. https://
forbes.com/sites/bernardm
arr/2024/05/06/why-ai-challenges-us-to-become-more-human/
… -
DistBelief Legacy: Model and Data Parallelism in Modern Neural Networks
By
–
I'm still a bit surprised that you think the system (DistBelief) that first combined large scale model- and data- parallelism with backprop-based training of very large neural nets is a "dead end", given that nearly every large-scale model that is trained today combines these
-
DistBelief convolutions development shared weights 2012
By
–
We supported shared weights from very early in DistBelief development, and we started adding support for various kinds of convolutions in March of 2012 (because convolutions are a very sensible idea for some things!), and you visited in July, I believe.
-
DistBelief Convolutions Backpropagation Training Before ICML 2012
By
–
DistBelief supported convolutions and end-to-end training with convolutions via backprop before the ICML 2012 paper appeared (even though that particular paper didn't use convolutions).
-
Mapping the Mind: Speech Production Understanding Advances
By
–
Mapping The Mind: Advances In Understanding Speech Production https://
forbes.com/sites/williamh
aseltine/2024/04/08/mapping-the-mind-advances-in-understanding-speech-production/?utm_source=dlvr.it&utm_medium=twitter
… #Innovation #ArtificialIntelligence #MachineLearning -
Matrix Classes and Neural Networks: From BPTT to Modern Frameworks
By
–
In the 90s I had to first implement matrix classes before I could even start coding neural nets – it was awful! I suffered so much with BPTT! Matlab did make things more efficient and was great until Python arrived. Leon Bottou and Yann Lecun pioneered SGD and torch style
-
Python and C++ in Computer Science Education Since 2005
By
–
There is a record. We started using python for teaching in 2008 (
https://
cs.ubc.ca/~nando/340-200
8/index.html
…) but the team was on if from around 2005. Of course, some people still preferred matlab (eg @sirbayes who still loves it ). Our MoE work (PAQ8) around 2008 was all on C++ because of