2 ms·

I follow Paul's argument about the consequences of the churning of low quality output by AI, but I think there's a second order effect that's more concerning. T

I follow Paul's argument about the consequences of the churning of low quality output by AI, but I think there's a second order effect that's more concerning. The ability to judge other people's knowledge of a subject area will become exceedingly rare and priceless. Unlike younger generations, who are growing up surrounded by AI-generated content, many of us older folks have had the experience of engaging directly with people and evaluating their competence. We developed a knack for quickly determining someone's skill level through just a few minutes of face-to-face conversation—a skill that was essential for navigating various life situations. Now that anyone can use AI to generate seemingly competent text, videos, and more, and as in-person interactions decline, the conditions that once allowed us to gauge competence are fading. I worry that in the future, no one—including AI trained on our outputs—will be adept at making these assessments. Those of us who take time to carefully compose arguments and revise them, as Paul suggests, will have a better handle on this, so that's a helpful consideration. I worry strongly about a future like that in Ideocracy[1], where nobody has a clue bout actually judge competence, and instead goes with the best sound bites. The one path out that I can see, and it's unlikely, is to teach the skill of explicitly tracking history, and reviewing how well someone predicted the future, over time. The explicit generation and curation of a reputation is part of that priceless nexus that they'll all be seeking in future generations, and yet it'll pale in comparison with the ability to size someone up in a few minutes of interaction. [1] https://www.imdb.com/title/tt0387808/ https://www.imdb.com/title/tt0387808/

hackable_sand 2y ago
Yes I been intentionally changing up my candor so that people Who get caught up in the structure, lose the message If u know you know
ndarray 2y ago
Judging other's competence will only be a problem if brain-computer interfaces become widespread before AI largely replaces the competent workforce, and my money is on AI replacement coming first. - Only a brain chip could make AI usage undetectable in practice. Without that you can tell if the person is checking his phone etc. Though you're right that an in-person interaction will be needed, otherwise there's no way of knowing what the other person is doing or if he's a real person at all... And since the latter problem (dead internet) will only grow, perhaps beyond the rectifiable, in-person communication will surely be in business again. - Once AI replacement of competent humans has reached a certain threshold, what do you stand to gain from testing a human's level thereof? Are you interviewing for "above AI" positions? If not, relying on AI will be as normal as relying on a calculator.
neilv 2y ago
> many of us older folks have had the experience of engaging directly with people and evaluating their competence. We developed a knack for quickly determining someone's skill level through just a few minutes of face-to-face conversation—a skill that was essential for navigating various life situations. I think I have a bit of this knack, in some areas, tempered by an awareness of some of my blind spots, but most people don't even claim to have this knack... As evidence from our own field: before the explosion of LLM cheating, we had the explosion of Leetcode hazing. Because, supposedly, good experienced software developers couldn't plausibly recognize each other just by talking with each other. So instead we whip out these douchetastic did-you-prep-for-this rituals. And afterwards you still have no idea what the other person would be like to work with (except that you now know both of you are willing to play fratbro nonsense games).

hackable_sand 2y ago

Yes I been intentionally changing up my candor so that people Who get caught up in the structure, lose the message If u know you know

ndarray 2y ago

Judging other's competence will only be a problem if brain-computer interfaces become widespread before AI largely replaces the competent workforce, and my money is on AI replacement coming first. - Only a brain chip could make AI usage undetectable in practice. Without that you can tell if the person is checking his phone etc. Though you're right that an in-person interaction will be needed, otherwise there's no way of knowing what the other person is doing or if he's a real person at all... And since the latter problem (dead internet) will only grow, perhaps beyond the rectifiable, in-person communication will surely be in business again. - Once AI replacement of competent humans has reached a certain threshold, what do you stand to gain from testing a human's level thereof? Are you interviewing for "above AI" positions? If not, relying on AI will be as normal as relying on a calculator.

neilv 2y ago

> many of us older folks have had the experience of engaging directly with people and evaluating their competence. We developed a knack for quickly determining someone's skill level through just a few minutes of face-to-face conversation—a skill that was essential for navigating various life situations. I think I have a bit of this knack, in some areas, tempered by an awareness of some of my blind spots, but most people don't even claim to have this knack... As evidence from our own field: before the explosion of LLM cheating, we had the explosion of Leetcode hazing. Because, supposedly, good experienced software developers couldn't plausibly recognize each other just by talking with each other. So instead we whip out these douchetastic did-you-prep-for-this rituals. And afterwards you still have no idea what the other person would be like to work with (except that you now know both of you are willing to play fratbro nonsense games).