>be given a set of rules>come up with novel and useful ideas that work with these existing rules >that were NOT already found in the initial set of rules example>give AI all math knowledge prior to isaac newton>an AI should be able to, unprompted: >identify the gaps>solve the gaps (create calculus) or>give AI all physics knowledge prior to einstein >an AI should be able to, unprompted: >identify the gaps>solve the gaps (create theory of relativity) current A""I"" cannot do anything remotely close to this so it should not be called AI
>>108972563>25x7you are the unintelligent
>all math knowledge prior to>all physics knowledge prior tothese are impossible things to teach and also foolish because the human understanding of physics and math before newton and the other enlightenment bros was fundamentally flawed.also if we applied such a test to a human being i believe most of humanity would fail because having knowledge and being able to utilize it are two seperate skills.
>>108972962>these are impossible things to teachthese are hypothetical examples to illustrate the concept of what I would expect from an "intelligence". this could be simplified down to giving basic rules on some specific subset of math and then having it figure out more. >if we applied such a test to a human beinginteresting how AI shills flip flop between "this is literally god" and "this is unfair a human cant even do this" also humans already HAVE been tested on this its called the history of every academic field ever
>>108972962Your arguments are shit.>>108972563Current AI is based on next token prediction. The one that goes after your prompt. And it does that because prompt is not the only context it has. Rest of it is stored in the model. Model itself is a compressed version of the Internet. Everything AI companies downloaded, purchased, stolen etc.Weights and biases format the model is stored in is basically a lossy compression. That's what the process of training is. Compressing the entire interwebs into mere gigabytes/terabytes.While you are correct when you say current AI is shit, you are a retard because in so many years you have not figured out what current AI actually is.Unless you proposed to create a model from a dataset completely cleansed of certain parts of Math and Physics and test it out like you just proposed. If so, I wonder how are you going to test it out, given the current situation on the market with the hardware prices.
