https://imgur.com/a/LfEbbL4
*using GPT-4o<p>1st prompt = describe the layout of this map.
//to confirm comprehension.<p>GPT's response:
"The layout of this map features a hierarchical tree structure. At the top level (level one), there is a single node that branches out into nine nodes on the second level. Each of the nine nodes on the second level then connects to six smaller nodes on the third level, making a total of 54 nodes on the third level.<p>The entire map is spread horizontally with a wide aspect ratio, ensuring all nodes are evenly distributed across the levels. The branches connecting the nodes are straight, creating a clean and symmetrical flow from the top node down to the lowest level. The design uses a consistent style for all nodes and connecting lines, making it easy to follow the hierarchy."<p>2nd prompt: recreate this map exactly 100% in every way;
The layout of this map should features a hierarchical tree structure. At the top level (level one), there is a single node that branches out into nine nodes on the second level. Each of the nine nodes on the second level then connects to six smaller nodes on the third level, making a total of 54 nodes on the third level.<p>The entire map is spread horizontally with a wide aspect ratio, ensuring all nodes are evenly distributed across the levels. The branches connecting the nodes are straight, creating a clean and symmetrical flow from the top node down to the lowest level. The design uses a consistent style for all nodes and connecting lines, making it easy to follow the hierarchy.<p>the only change is to make each square a purple isometric shape.<p>// Is there a better way?
There are a lot of things it can't do. Try making it sort a list in some order that isn't repeated repeatedly in the training set (e.g. not U.S. states in alphabetical or reverse alphabetical order)<p>You can even ask if it is able to do the task and it will tell you that it can, but it can't, any more than it can help you make a neutron bomb. The difference is it will tell you it can't make a neutron bomb.<p>LLMs shine at problems that involve linear relationships where you go mostly from one end to the other, like language translation. Give them something that involves complex relationships and they quickly fall apart.