A jargon-free explanation of how AI large language models work
arstechnica.com
arstechnica.com
The transformer figures out that wants and cash are both verbs (both words can also be nouns). We’ve represented this added context as red text in parentheses, but in reality, the model would store it by modifying the word vectors in ways that are difficult for humans to interpret. These new vectors, known as a hidden state, are passed to the next transformer in the stack.
If you want to learn how llms work, I'd suggest "GPT in 60 lines of numpy": https://jaykmody.com/blog/gpt-from-scratch/
It's more concise and doesn't resort to analogies.
Your alternative site serves an entirely different audience/purpose.
I'm surprised to see this submission did relatively poorly. HN has seen so many trivial "what is tokenization" articles. This Ars article is actually substantive and gives a very real notion of how LLMs work via their components/successors. A++
I think it was too long and too "deep" but at the same time geared towards a nontechnical audience and trying to explain by analogy, so it's going to lose people that don't already understand and annoy people who do.
What I like about the link I posted is that it's approachable to anyone who understands python and some basic preliminaries. It provides understanding to an audience that's potentially capable of understanding. The Ars article tries to do to much and I expect will rarely succeed in giving someone who can't understand my article (I posted it I didn't write it) an understanding of transformers.