Learning from successful founders & companies often leads us into traps. We pick bad examples and learn the wrong lessons. Worse, even knowing that we make these mistakes doesn't seem to help much. I discuss the research, and some potential solutions: https://
open.substack.com/pub/oneusefult
hing/p/when-survivorship-bias-meets-superstitious?r=i5f7&utm_campaign=post&utm_medium=web
…
RESEARCH
-
Survivorship Bias and Superstition in Startup Learning
By
–
-
Neural bandwidth: All brain nerve input fits through single Ethernet cable
By
–
All nerve input to the brain combined could run down a single Gb Ethernet cable, and only an insignificant fraction is retained in any way.
-
DAAM: Interpreting Stable Diffusion Cross Attention
By
–
What the DAAM: Interpreting Stable Diffusion Using Cross Attention Tang et al.: https://
arxiv.org/abs/2210.04885 #ArtificialIntelligence #DeepLearning #MachineLearning -
AI vs Machine Learning vs Deep Learning: Understanding the Differences
By
–
#AI Vs #MachineLearning Vs #DeepLearning
Via @kenovy_it Artificial Intelligence aids in creating intelligent machines.
ML is a subset of AI used to build AI-driven applications.
DL is a subset of ML, it uses massive sets of #data and complex algorithms to train models. -
Training Degeneracy in Rectangular Weight Matrices Without Hadamard
By
–
This is the training degeneracy that @jiawzhao is talking about if you only do identity initialization and not Hadamard for rectangular weight matrices
-
ZerO Initialization and Hadamard Transforms for Rectangular Matrices
By
–
And an overview of how ZerO initialization can overcome degeneracy by using Hadamard transforms for rectangular weight matrices
-
ZerO Initialization: Neural Networks with Only Zeros and Ones
By
–
ZerO Initialization: Initializing Neural Networks with only Zeros and Ones A fully deterministic initialization scheme which sets the weights to only 0s and 1s can achieve SOTA on various datasets including ImageNet. Maybe random weights are unnecessary. https://
openreview.net/forum?id=1AxQp
KmiTc
… -
Data Quality Critical: Flawed Input Ruins AI Systems
By
–
Y'all know my #AI trinity: chicken (algorithm), eggs (data), & bacon (objective). Here's an example of rotten eggs: https://
bit.ly/3WUGCGs You need to know the data going into your AI for BOTH training and input. This was flawed automated input & could have easily been expected. -
Big IoT Data Analytics Through Fog Computing Technology
By
–
Big #IoT #DataAnalytics via Fog Computing http://
bit.ly/35ubEvj #Analytics #MachineLearning #AI #Python #Rstats #5G #FutureOfWork #flutter #java #IIoT #BigData #Cloud #Edge #SmartCity #CES2023 #FutureOfWork #100DaysOfCode #SDGs #Sustainability #DataScience #Automation -
Future Fund crisis threatens NeurIPS ML Safety workshop grants
By
–
"…there are many committed grants that the Future Fund will be unable to honor." The #NeurIPS2022 ML Safety Workshop (
https://
neurips2022.mlsafety.org) had an eye-watering $100k of awards for best papers and AI risk analyses. Given all the turmoil, I presume these are getting the axe?