Your point about all three conditions needing to fail simultaneously is the strongest counterargument I've seen. But it also cuts the other way. Nobody modeled the scenario where two of the three fail and the third holds. Or where they fail sequentially instead of simultaneously. The real gap in this whole debate isn't optimism vs pessimism, it's that everyone's arguing about one scenario when the actual range of outcomes is barely being explored.
This is a better version of my own argument than I wrote. You're right that the branching tree has more than two endpoints, and nobody is mapping the middle paths. The sequential failure point is especially sharp. Diffusion buys time. But time only matters if something useful happens inside it and we have considerable evidence that institutions tend to waste exactly the kind of time that diffusion provides. I've written elsewhere about the temporal mismatch between AI development (months), institutional response (years), and human adaptation (decades), and what that mismatch reliably produces: solutions to yesterday's problems arriving just in time to miss today's. The Citrini piece and the response to it have convinced me that the next serious piece of writing on this needs to map those middle scenarios explicitly. Not crisis or dividend, but the messy paths where one condition holds, another fails, and the third degrades slowly enough that nobody notices until it's structural. Thank you for pushing the argument forward.
That temporal mismatch framing is exactly right. Months vs years vs decades, and everyone arguing as if they're all on the same clock. Someone should map those middle paths properly.
Still waiting for someone, anyone, to explain the new roles and new tasks that will emerge as algorithms decimate knowledge work. It's already been 3 years and so far there's no sign of this bright future for the white colllar worker, especially since the machines are being explicitly tailored to get better and better at managing themselves. As far as I can tell, the only humans in the loop are the slave wage workers overseas (and increasingly in the west as more laid off tech workers resort to data annotation work).
And it *is* still prompt engineering, no matter how much certain techbros try to dress it up. Ordering outputs from a vending machine is the illusion of work.
Your point about all three conditions needing to fail simultaneously is the strongest counterargument I've seen. But it also cuts the other way. Nobody modeled the scenario where two of the three fail and the third holds. Or where they fail sequentially instead of simultaneously. The real gap in this whole debate isn't optimism vs pessimism, it's that everyone's arguing about one scenario when the actual range of outcomes is barely being explored.
This is a better version of my own argument than I wrote. You're right that the branching tree has more than two endpoints, and nobody is mapping the middle paths. The sequential failure point is especially sharp. Diffusion buys time. But time only matters if something useful happens inside it and we have considerable evidence that institutions tend to waste exactly the kind of time that diffusion provides. I've written elsewhere about the temporal mismatch between AI development (months), institutional response (years), and human adaptation (decades), and what that mismatch reliably produces: solutions to yesterday's problems arriving just in time to miss today's. The Citrini piece and the response to it have convinced me that the next serious piece of writing on this needs to map those middle scenarios explicitly. Not crisis or dividend, but the messy paths where one condition holds, another fails, and the third degrades slowly enough that nobody notices until it's structural. Thank you for pushing the argument forward.
That temporal mismatch framing is exactly right. Months vs years vs decades, and everyone arguing as if they're all on the same clock. Someone should map those middle paths properly.
Still waiting for someone, anyone, to explain the new roles and new tasks that will emerge as algorithms decimate knowledge work. It's already been 3 years and so far there's no sign of this bright future for the white colllar worker, especially since the machines are being explicitly tailored to get better and better at managing themselves. As far as I can tell, the only humans in the loop are the slave wage workers overseas (and increasingly in the west as more laid off tech workers resort to data annotation work).
And it *is* still prompt engineering, no matter how much certain techbros try to dress it up. Ordering outputs from a vending machine is the illusion of work.
every word of this is ai. pathetic