It seems obvious to me by now, that it's not LLMs as tools that are flawed but rather the interface through which we are forced to interact with them, or, eventually, the human using the tool. The former influences the latter, making it more important than it's currently given credit for
3 months ago