Posts
Why I am writing in English and not Arabic.
There is a guilt that comes with it. shame on me for writing in English while being Arab by mother and father. I have never fled to a western or European country. I was not obsessed with western movies, in fact I barely watch movies at all. I accidentally watched two movies in the past two years and that was it. yet I am still influenced by the west. that is simply the reality of the world we live in today. influence no longer requires migration. it travels through knowledge, through the internet, through science, through code.
I remember once posting on Facebook about how I dislike people who compliment me , and I wrote something along the lines of: if you want to be a good friend of mine, do not give me compliments.
Recently, I’ve been working on the reproducibility of machine learning research. In practice, this meant doing far more infrastructure work than actual machine learning modeling. I found myself digging into operating systems, containers, storage, and hardware details. and suddenly, I started questioning everything, not because it wasn’t interesting, but because I was drifting away from the kind of problems that originally made me fall in love with machine learning.
do things badly
As the years pass by, I find myself adapting to different reasons to keep living, and to share, I’m not as clear as I used to be about why I should be posting something publicly, but looking back over the years, I recall how writing started. How reasons changed.
Reproducibility is the corner stone of science
Somewhere between building a model and publishing the results, something gets lost. not the results themselves, those make it into the paper. what gets lost is everything that would allow someone else to arrive at the same place.
This is the reproducibility problem in machine learning, and it is more widespread than most people admit.
In 2016, the journal Nature reported that around 70% of researchers had failed to reproduce another researcher’s results , and 50% had failed to reproduce their own. machine learning is no exception. a study that analyzed 400 papers from top AI conferences found that only 6% shared code, roughly 33% shared test data, and 54% shared nothing more than a pseudocode summary of their algorithm. not the environment. not the hyperparameters. not the exact version of the library that made it work. just a rough sketch of the idea.
And I remember telling one of my closest friends, hey just out of the blue it hit me I turned to him and I said: you now, you are one of those friends if I was really in the shit I Would call you, and his reaction, is he didn’t say thank you, he said, I’d be mad if you didn’t, when you don’t call me and say I’m in the mud, you deny me the opportunity, the honor of setting in mud with you. - Simon Sinek