Is Attention All You Really Need?

On the cover: MLA Architecture. Credits: Welch Labs Large language models generate text one token at a time by taking all the tokens that came before them as input. They are the classic autoregressive models after all. At step t, the attention fo...

Markov Chain - The connection between Nuclear Bombs, Google Search, and Perplexity

On the cover: A Nuclear Bomb Explosion Recently I watched this video by Veritasium talking about Markov Chains. I really liked the examples he used to explain the usage of Markov Chains, ranging from Nuclear Bombs to Google Search and Perplexity ...

HNSW - Finding Needles in Vector Haystacks

On the cover: HNSW graph depiction. Credits: Marqo Imagine a city with 10 million coffee shops. You (the customer) have a very specific taste vector: medium roast, cozy seating, quiet music. Now how do you go about finding the best coffee shop? ...

Padlocks to Prime Numbers - RSA Encryption and SSH

On the cover: RSA Encryption If you’ve ever shopped online, sent a private message, or connected to a secure website, you’ve probably used RSA encryption without even knowing it. RSA was founded by three mathematicians Rivest–Shamir–Adleman (RSA)...