[danielmarkbruce]

I have some sense that it would work better than most CEOs. It would require someone reasonably smart to build the system and keep it fed with the right data etc, but the amount of decision making and additional context as to why decisions are made etc would be valuable, especially where CEOs are a bottle neck (which is... a lot of places)

[saulpw]

If you have a strategy and need plans, maybe you could use an AI (probably not LLMs). If you need a strategy, maybe you can feed high-level goals into an AI and it could come up with one (though it will have difficulty thinking outside its box). But the primary value of a CEO is being the definitive source of high-level goals and systemic values. An AI is not going to come up with those, except by mimicking the stated goals and values of other CEOs.

It's easy to say "the high-level goal is maximizing profit" but as we all lament daily, having that as a goal is both vapid and only functional in the short-term.

[danielmarkbruce]

Why can't it come up with those?

[saulpw]

I mean sure, it can make things up just like any fake CEO can. I can ask ChatGPT for a list of goals and values right now and they might even sound reasonable.

But there's no discernment, no conviction. No actual 'seeing', just words and concepts thrown together in an attempt to pass muster. You can argue that's all that humans are doing, and to be sure, there are many people who are faking it like this all the time, sometimes through their entire life, and even in high-level positions. One of them is even running for president right now. But when you compare one of these fakers with someone who has actual principles and values and meaningful goals, there's clearly a bright line difference. And there's no way the current round of AI/LLMs could have that, no matter how much they sell it or say that's around the corner.