Konrad Kording

5 Followers
283 Following
96 Posts
@Penn
UPenn Prof, CIFAR Learning and Machines co-director, Community for Rigor (C4R), deep learning, brains, #causality, scientific rigor, Bayes, Machine Learning, http://neuromatch.io, Transdisciplinary optimist, Dad, Collaborator, Loves outdoors, hiking, skiing 🦖
In statistics, measurement is hard. Comparing two measurements is twice as hard. Comparing two comparisons is twice as hard again.
@PessoaBrain @NicoleCRust @charanranganath @DrYohanJohn @dbarack @SussilloDavid @kordinglab But if we consider reservoir computing as “computation through dynamics” then our results show each node is somewhat involved, e.g., from 36 nodes only 2-3 of them have “zero contributions”. At least to our simple task and in our very small model.
@kordinglab I knew it! After all, a prior (or prioress) is an ecclesiastical title for a superior in some religious orders https://en.wikipedia.org/wiki/Prior
Prior - Wikipedia

@seeingwithsound a careful analysis showed that this bias affects all aspects of science but not those that like Bayesian models. Their models are actually so amazingly useful and wonderful. Their experiments are guided directly by the almighty.
Today I learned about a concept that I did not have a name for: "Pet model" bias. There are ideas we love. That taints our choice of questions, experimental design, data analysis, and reporting.
Biased AI systems make humans biased: https://osf.io/c4e7r
So Sfn was really great. I see our five coming years of community for rigor pretty clearly ahead of me.
I wish dinosaurs knew how famous they are now