All reasoning involving fuzzy, ill-defined categories is suspect. Arguments without quantifiers (all, most, some) are equally suspect. Fully agreed with these examples.
My only disagreement is with trying to rank them—these are all in the category of “worst”. Trying to reason about an individual based on fuzzy group membership (or based on other individuals via some shared attributes) is pretty horrific.
In the direction you’re pointing, “whataboutism” is pretty rampant in political conversations. “yeah, X has problems, but Y and Z had vaguely-similar problems”. No known solution yet, except to recognize that your interlocutor isn’t TRYING to come to consensus or change their mind. I suspect what you’re seeing about ASI risks is similar—it’s not that they really want to estimate probabilities, they just want to stop discussing it.
Note that they may be right, even if argued badly—humans are pretty horrible to each other, and likely will destroy ourselves even without ASI. We’ve been well on track for the last few decades toward a civilizational collapse due to environmental, political, and supply-chain fragility. Stronger AI may accelerate it. Or it may slow it. Or it may do something unexpected entirely. But it’s not guaranteed massively worse than the status quo ante.
My only disagreement is with trying to rank them—these are all in the category of “worst”.
I don’t disagree—“second-worst” here is more or less a rhetorical flourish to make clear the connection with Scott’s well-known essay on the noncentral fallacy.
Trying to reason about an individual based on fuzzy group membership (or based on other individuals via some shared attributes) is pretty horrific.
Agreed. I’ve been reading Lakoff bit by bit over the past few months and I’m now about halfway in. It’s opened my eyes tremendously to how much bad argumentation (whether deliberately obtuse, or well-intended but sloppy) relies on exploiting the fuzzy nature of categories.
All reasoning involving fuzzy, ill-defined categories is suspect. Arguments without quantifiers (all, most, some) are equally suspect. Fully agreed with these examples.
My only disagreement is with trying to rank them—these are all in the category of “worst”. Trying to reason about an individual based on fuzzy group membership (or based on other individuals via some shared attributes) is pretty horrific.
In the direction you’re pointing, “whataboutism” is pretty rampant in political conversations. “yeah, X has problems, but Y and Z had vaguely-similar problems”. No known solution yet, except to recognize that your interlocutor isn’t TRYING to come to consensus or change their mind. I suspect what you’re seeing about ASI risks is similar—it’s not that they really want to estimate probabilities, they just want to stop discussing it.
Note that they may be right, even if argued badly—humans are pretty horrible to each other, and likely will destroy ourselves even without ASI. We’ve been well on track for the last few decades toward a civilizational collapse due to environmental, political, and supply-chain fragility. Stronger AI may accelerate it. Or it may slow it. Or it may do something unexpected entirely. But it’s not guaranteed massively worse than the status quo ante.
I don’t disagree—“second-worst” here is more or less a rhetorical flourish to make clear the connection with Scott’s well-known essay on the noncentral fallacy.
Agreed. I’ve been reading Lakoff bit by bit over the past few months and I’m now about halfway in. It’s opened my eyes tremendously to how much bad argumentation (whether deliberately obtuse, or well-intended but sloppy) relies on exploiting the fuzzy nature of categories.