> Chatgpt is way better at being persuasive than it is at being correct.
I get the sentiment, but I doubt this is statistically true. Of course it makes mistakes, and certain kinds more than others, but ChatGPT’s recall is very good.
> It is probably already netter than you at convincing them of stuff.
I don’t think I agree, but I’m not clear what you mean. What do you mean? Care to unpack it?
Here’s what I mean when I say I disagree:
* ChatGPT, on the whole, doesn’t press an agenda. While it does have baked-in opposition to violence, hurting people, sabotage, bombs, etc, it does not (best I can tell) have a values system to weigh competing interests in a workplace.
* Agency. This one may be obvious, but I think it needs to be stated. If you want ChatGPT to weigh in on a decision, you have to ask. You have to frame the question. An employee/contractor can come in with their own agenda and present a recommendation.
* Much of the time (I’m not sure how often, this is an empirical question), when interacting with something like ChatGPT, people convince themselves. They frame the questions. They read answers. They stop at some point, perhaps when convinced, perhaps when the response seems incorrect, and/or perhaps when they run out of time or interest. Along the way, they can push back, asking follow-up questions or disagreeing. Of course, ChatGPT is presenting information and that information may be persuasive. But what “get’s in” to the human brain varies a lot based on what that person is receptive to.
* Of course, a significant portion of Americans demonstrate limited critical thinking skills, for large portions of their waking life. One might debate this point, i.e. saying that sometimes just “going along with something” is a rational strategy to conserve energy, avoid conflict, fit in with a group, and so on. Whatever is going on, though, it goes appear that for many people, belief formation is not primarily driven by a quest for truth in the classical liberal sense. Such people often seek “advice” from ChatGPT not with the intent to question it, but rather to reduce the time and effort that thinking and researching requires.
I get the sentiment, but I doubt this is statistically true. Of course it makes mistakes, and certain kinds more than others, but ChatGPT’s recall is very good.
> It is probably already netter than you at convincing them of stuff.
I don’t think I agree, but I’m not clear what you mean. What do you mean? Care to unpack it?
Here’s what I mean when I say I disagree:
* ChatGPT, on the whole, doesn’t press an agenda. While it does have baked-in opposition to violence, hurting people, sabotage, bombs, etc, it does not (best I can tell) have a values system to weigh competing interests in a workplace.
* Agency. This one may be obvious, but I think it needs to be stated. If you want ChatGPT to weigh in on a decision, you have to ask. You have to frame the question. An employee/contractor can come in with their own agenda and present a recommendation.
* Much of the time (I’m not sure how often, this is an empirical question), when interacting with something like ChatGPT, people convince themselves. They frame the questions. They read answers. They stop at some point, perhaps when convinced, perhaps when the response seems incorrect, and/or perhaps when they run out of time or interest. Along the way, they can push back, asking follow-up questions or disagreeing. Of course, ChatGPT is presenting information and that information may be persuasive. But what “get’s in” to the human brain varies a lot based on what that person is receptive to.
* Of course, a significant portion of Americans demonstrate limited critical thinking skills, for large portions of their waking life. One might debate this point, i.e. saying that sometimes just “going along with something” is a rational strategy to conserve energy, avoid conflict, fit in with a group, and so on. Whatever is going on, though, it goes appear that for many people, belief formation is not primarily driven by a quest for truth in the classical liberal sense. Such people often seek “advice” from ChatGPT not with the intent to question it, but rather to reduce the time and effort that thinking and researching requires.