Sure but then you have added a very large layer of complexity just to turn your operating system into (a very inefficient) prompt box.

It couldn't learn in any meaningful way unless you continuously train the models and prompt it to be aware of its suggested direction.

At the end of the day you aren't getting an LLM to replace the humans that select the data they are trained on.

Look to Machine Learning. The overlap is where AI shines.

Reply to this note

Please Login to reply.

Discussion

Here is some more info in the ballpark of "LLM operating systems" but there is still the major caveats I mentioned above

https://news.ycombinator.com/item?id=43294974

https://news.ycombinator.com/item?id=37901902

https://news.ycombinator.com/item?id=37894403