The art AIs have artistic talents, but don't have the LLM's abilities in understanding complex structure. It's really a different discussion. Take, for example, what GPT-4 came up with when I described
yesterday's xkcd to it and asked what it would take to simulate it. There's no place this reply could have been cut-and-pasted from, since it's new. It could not have been written without it actually understanding a number of concepts.
GPT-4:
To simulate this scenario, you'd need to consider a few factors:
1. Precise Physical Model: You'd need a detailed model of Earth's tectonic plates, including data about their current positions, velocities, and directions of movement. This information could help you understand how quickly the Atlantic Ocean is expanding.
2. Dynamic Text Editor: This software would need to automatically adjust text as the "page size" (i.e., the surface area of the Atlantic Ocean) increases. It would have to handle wrapping text around the coastlines and adding additional words as the ocean expands. The text editor would also need to accommodate an immense amount of text, since the Atlantic Ocean covers over 41 million square miles.
3. Word Generation: The system would need a way to generate or add new words as the Atlantic Ocean expands. This could involve generating random words, copying from an existing text, or using an AI language model to create coherent and contextually appropriate words.
While it would be technically challenging to implement such a system, it's an interesting thought experiment that encourages us to think about geological processes, software design, and the physical scale of Earth in new and unique ways.