Tobi Lehman's blog

# Word embeddings and semantics

Can a machine understand what a word means? Right now machines routinely correct spelling and grammar, but are pretty useless when it comes to semantics. Search engines are an exception, they have a rudimentary understanding of what words mean. One of the ways this can work is explored in Tomas Mikolov’s 2013 paper on word embeddings. Word embeddings are mappings from sets of words to vectors, such that the distances between the vectors represent the semantic similarity of the words. Read more...

# Thing explainer in emacs

This is my first major mode for emacs. It was inspired by Randall Munroe’s Thing Explainer and Morten Just’s editor that doomed humanity. The concept is simple, restrict your vocabulary to the 1000 most common words. If you can explain something using this reduced vocabulary, then you really understand the topic. This is a decent test of understanding because it’s easy to learn a word, and even use that word in the right context, but still have no idea how it relates to other things. Read more...

# Dockerize all the things

Okay, maybe not all the things, but the things that tend to litter your filesystem with libraries, dependencies and other crap that increases the chances of a conflict. If you are not familiar with Docker, check it out, then come back here and troll my flaming fanboy drunk on kool aid ravings in the comments. I write a lot of ruby and python code, and I also use a lot of ruby/python and javascript code other people wrote as well. Read more...

# New largest Mersenne prime found

The largest prime number was just found, and it is a Mersenne prime, or a prime number of the form 2n - 1. As of January 2016, 274,207,281 - 1 is the largest known prime. Primes are interesting in their own right, but they also are indispensible in cryptography. We’ve known since Euclid that the number of primes is infinite, but it is still an open problem whether the number of Mersenne primes is infinite: Read more...