# Pattern Matching vs Understanding You can make an argument that you must distinguish *pattern matching* from *understanding*. Scene analysis is completely different from computer vision. Scene analysis is understanding the interactions between objects-if I were to throw this ball at the floor, what would happen? If that were to be the case, the argument goes, the information is *not in the data*. A counter argument would be: If the information is not in the data, then how do humans learn it? You may say that we have some *innate* knowledge given to us at birth. But then again you must ask, how was the innate knowledge learned? Evolution is also just a learning algorithm, the information had to get there some how. Where did it come from? Why don’t we all think gravity goes up instead of down? It is because we are exposed to it. It shapes out evolutionary fitness. If a species evolved that thinks gravity goes up instead of down, that would be lower fitness. That is a training signal. It is an exchange of *Shannon mutual information* between the environment and the learning algorithm. --- Date: 20230114 Links to: [AI MOC](AI%20MOC.md) Tags: #review References: * [#031 WE GOT ACCESS TO GPT-3! (With Gary Marcus, Walid Saba and Connor Leahy) - YouTube](https://youtu.be/iccd86vOz3w?list=PLwcClAaLqrJy4wrVTB43HDlnzW-d20fUt&t=6175)