Staying on top of research is hard. New research doesn't have to be consumed immediately, give it a year if it's still relevant and check citations. We need more peer and quality feedback. Don't drink from firehose. #arxiv
A prediction E(Y| X, E, T) of outcome Y given covariates X is best for that environment E at time T, even if X is constant, E is dynamic and will change with T. Whether E and T should and can be in X is an open question. E and T are usually overlooked (drift).
@easyJet flew from athens, your checkin had 2 staff, hence people had to run to not miss the plane, which then stood on runway for 45 mins, you created unnecessary stress for customers
Product recall: Can we return Johnson back to Eton and Oxbridge? The product was not as good as promised. Did not meet quality standards. Not even a good speaker (peppa pig).
I think AI in 2030 will be way more busy playing computer games than help humanity, those games are literally made for them, the rewards are so much clearer :)
Christensen's idea of the "job to be done" (customer need) is a great one, and overlooked by some new businesses who start with the tech and not the need #marketing
Instead of a GAN for continuous data you can just extract MultivariateNormal(Mu, Sigma) from the data, and then generate (simulate) your private synthetic data, that maintains variable relationships #syntheticdata
@omid9 I heard you on the LBC podcast, I miss you being on UK TV. I could not find the English version of the Persian talk show you did. #bringbackomid