#biology #computing #ai # Zero shot [[Biology/Matt Ridley|Ridley]], when talking about the relationship between the favourite evolutionary biology couple, nature and nurture, said: >These and other mechanisms in the brain are "innate" and highly specific to their task, but they are probably perfected by exposure to examples. No general-purpose induction here. The context was human vision. Current [[Artificial intelligence|AI]]s are able to do what is called [[Zero shot learning]]. It seems animals do it too? Our training is natural but we run inference with our nurture context. Let's say I'm in conversation, the ability to say words by moving parts of my head, is natural. The general semantic structure of any language is also imprinted in my [[Genome|genome]], nature. But when I have to answer a sentence of someone else, I basically do a [[GPT3]] inference: ``` Bob: blabla Me: bloblo Bob: blibli Me: ``` Given the conversation context, I predict the future words, and predicting successfully is understanding, is [[Philosophy/Rationality/Intelligence]]. # Few shot We also do [[Few shot learning]]: Another example, take a child who didn't learn math yet. You teach him additions then give him a substraction and ask him $2-1=$ He will likely generalise his minimal math understanding and answer 1. Inference will be 90% nature (world model) and 10% zero shot from nurture as context (addition [[Philosophy/Epistemology/Knowledge|knowledge]]).