Intro
Lately I’ve been thinking a lot about intelligence, what is exactly intelligence, and what we would need in order to create “Real” artificial intelligence. In this post I’m trying to analyze my reasoning process and make conclusions about the current state of AI.
The two side of the same coin
There is two things that goes into intelligence.
- The first is knowledge, these are the building blocks of reasoning.
- The second is an innate ability to think of related items and use them in an “intelligent” manner.
Knowledge is a set of known facts, like Water is wet, fire burns, or the sum of the angles in a triangle is 180º We can manipualte these facts to creates new informations, this is where the second part kicks in. Tut this is not enough, in order to be intelligent, we need that “spark” of pure genius when facing a problem and saying “Hey, but what if …” and comming up with an nice solution.that innate ability is not really defined, nor has a name, we can call it creativity, or imagination but IMO it is still too restrictive. Together these two things create the basis of intelligence.
LLMs and Reasoning.
LLMs get their facts from a pre-training on a large set of data, and to achieve the second one steps of intelligence, they brute force throught all the items that are probably closely related to the previous text. The insight in this is that it’s highly innefective