Differences in AI vs. Human learning and how machines can get good at it
date
Feb 6, 2023
slug
ai-vs-human-learning
author
status
Public
tags
Theory
Unedited
summary
type
Post
thumbnail
updatedAt
Feb 12, 2023 10:14 AM
Current LLMs are fed the whole conversation as input (work in the “language” space). Machines are good at memory. Humans don’t remember every single word previously said in a convo. Instead they form a representational sketch of what was said, the big ideas (work in the “idea” space). Maybe look into a model that can do that?
Early human learning is about generalization and association. Association: baby is shown mother while word “mom” is spoken. Generalization: rules are abstracted from the specific examples, e.g. “He eats” ⇒ <noun> then <verb>.
I feel like a big possible reason LLMs such at system 2 thinking is that they can’t actually interact with the world and get feedback.