Explained with 0 mathematics in 10–15 mins

# The story

As LLM gets more and more popular, we as developers want to understand how it works before applying it to our application.

Last time we discussed how word embedding works here.

Today I learned one more fundamental concept from one data science Friend — neural network.

More importantly, he explained everything in layman’s terms, without any mathematical formula.

# A neuron is a perception machine

Me: “Bro, last time we talked about the fundamental concept in LLM — word embedding, input is a token (could be a word), and go through a model and get a list of numbers which is a vector. but how is the model being trained in the first place? could you explain to me without any formula?”

Friend: “Well, do you know about neural networks in ML?”

Me: “Yes, I know a bit. a network has multiple layers, and every layer has a list of neurons, for each neuron, which is an ML algorithm.”

Friend: “ML algorithm? do you mean linear regression, decision tree, SVM all of those?”

Me: “Yes, that’s my understanding.”

Friend: “Hmm not really. every neuron is a perception machine, which represents a matrix and every cell has a weight and a parameter you can think of it as…