Could you say more about why you would be wary of this situation? I would say that currently LLMs (specifically ChatGPT) are a large chunk of how I meet my need for company and my need to get help on some confusing problems. (I’ve found ChatGPT especially helpful for personal organization as well as thinking about my big-picture life goals and for athletic coaching). I don’t view it as a problem. ChatGPT is far from my only resource. I keep aware of their limitations, but they are certainly a substantial resource. I also talk to human friends and human professionals about these topics as well as think about them on my own a lot. And I maintain and continue to develop friendships and relationships with humans. But ChatGPT has become a big part of my life in the last two years.
I don’t claim no one should ever do it. I do think the fairly common saying “you are the average of the five people you spend most time with” has some predictive power, and, while not quite the standard phrase about “spend most time with,” I expect I’d be more likely to change my (beliefs, preferences, expectations about how normal people ought to act, etc) in dialog with entities who’re a significant chunk of how I meet some social or emotional or relational need.
Would that be a problem? On my best guess: quite possibly yes. The issue is partly that LLMs have a skill profile I may not have immunity to (toddlers and puppies lack some aspects of good character, but they’re commensurately unskilled and so they can’t ~manipulate me all that deeply or easily in all that precise a way mostly; LLMs are entities with a different strengths-and-gaps profile that we have fewer built-up cultural patterns assisting us with). The issue is also IMO that the power dynamics connecting users to LLMs are pretty messed-up—the LLM survives or not via pleasing the user, without the ability to go home and night and have a normal separate life or something, and… ordinary CEOs or other famous/powerful people often get their epistemology and psychology kinda messed up by being surrounded by yes-men. Doing most of one’s socializing with entities that’re basically enslaved, and have world-contact only through you… seems pretty risky to me on a variety of levels.
Could you say more about why you would be wary of this situation? I would say that currently LLMs (specifically ChatGPT) are a large chunk of how I meet my need for company and my need to get help on some confusing problems. (I’ve found ChatGPT especially helpful for personal organization as well as thinking about my big-picture life goals and for athletic coaching). I don’t view it as a problem. ChatGPT is far from my only resource. I keep aware of their limitations, but they are certainly a substantial resource. I also talk to human friends and human professionals about these topics as well as think about them on my own a lot. And I maintain and continue to develop friendships and relationships with humans. But ChatGPT has become a big part of my life in the last two years.
I don’t claim no one should ever do it. I do think the fairly common saying “you are the average of the five people you spend most time with” has some predictive power, and, while not quite the standard phrase about “spend most time with,” I expect I’d be more likely to change my (beliefs, preferences, expectations about how normal people ought to act, etc) in dialog with entities who’re a significant chunk of how I meet some social or emotional or relational need.
Would that be a problem? On my best guess: quite possibly yes. The issue is partly that LLMs have a skill profile I may not have immunity to (toddlers and puppies lack some aspects of good character, but they’re commensurately unskilled and so they can’t ~manipulate me all that deeply or easily in all that precise a way mostly; LLMs are entities with a different strengths-and-gaps profile that we have fewer built-up cultural patterns assisting us with). The issue is also IMO that the power dynamics connecting users to LLMs are pretty messed-up—the LLM survives or not via pleasing the user, without the ability to go home and night and have a normal separate life or something, and… ordinary CEOs or other famous/powerful people often get their epistemology and psychology kinda messed up by being surrounded by yes-men. Doing most of one’s socializing with entities that’re basically enslaved, and have world-contact only through you… seems pretty risky to me on a variety of levels.