A real world analogy, concerning the adoption speed of AI in all fields, is the adoption rate of autonomous machines in transportation. When statistical analysis proves autonomously driven cars can provide safety that's equal to the average of human driven cars, the above average human driver will still provide a safer driving experience than the AI driver. Nearly all human drivers think they have above average driving skills, and close to half of them are wrong.
Only when autonomous vehicles drive as good as the best human drivers, will most humans trust full autonomous driving. The road in-between will be filled with human drivers using autonomous features at ever increasing rates, for the simpler tasks, thereby slowly teaching AI how to drive as good as the best humans do. In other words, something we thought would take a few years to achieve (widespread AI replacement of human agents) will actually take many decades.
Thank goodness this is the case. Humans are critically important to making stuff happen, and will be, far into the future. Slow your roll, much money and happiness will be lost due to over estimating the speed potential of this transition, IMHO.
Totally agree, humans being critically important is the entire thesis of our work - and the "slow roll" pun is perfect in an article using convection cells as its analogy, lol!
Good catch. I just don't get why you think human inputs will be "scarce" when logic says they will be abundant. Also, why do you think capital will be abundant when logic says it will be scarce?
To clarify, not “human input”, competent human attention. The idea that is as machines become capable of offloading routine cognition as a commodity, the remaining human roles become the binding constraint. The impact of this change at scale literally shifts the macro-economic balance between labor and capital. Have a look at our first paper “Is Human Attention Still Necessary?” which outlines this argument and the potential implications carefully.
I did read it, carefully, and common sense tells me that labor will become less important in relation to productivity while capital becomes more important. In other words expenditures aimed at materials will consume more of the pie than expenditures aimed at payrolls ..... for a any given production level.
Labor will cost relatively less as it becomes LESS SCARCE (mundane cognitive work is becoming more automated) while capital will cost relatively more as it becomes MORE SCARCE (upfront costs of automation rather than extended costs of labor).
Future reality appears to be the opposite of what you envision. If your corporate model guarantees loss of company control to those that fund its foundation (critically important) and promises control to those who can be easily and inexpensively hired, your company might never get off the ground.
You differentiate between "human input" and "competent human attention" but they are one and the same. The availability of "competent humans" will increase as their jobs become automated, right?
Appreciate the back and forth, Blain, and I mean it when I say disagreement is welcome here.
One thing for the record, since it's the actual crux: the paper doesn't argue that human input becomes scarce. It separates routine cognition, which is getting commoditized, from non-routine judgment, which isn't. Your case treats those two as the same thing, and that's the reason we keep talking past each other. Freeing people from routine work doesn't make judgment abundant. It makes judgment the binding constraint. That's the whole claim, start to finish.
We see it differently and that sits fine with me. I'm going to leave it here and get back to building. Thanks for reading closely.
A real world analogy, concerning the adoption speed of AI in all fields, is the adoption rate of autonomous machines in transportation. When statistical analysis proves autonomously driven cars can provide safety that's equal to the average of human driven cars, the above average human driver will still provide a safer driving experience than the AI driver. Nearly all human drivers think they have above average driving skills, and close to half of them are wrong.
Only when autonomous vehicles drive as good as the best human drivers, will most humans trust full autonomous driving. The road in-between will be filled with human drivers using autonomous features at ever increasing rates, for the simpler tasks, thereby slowly teaching AI how to drive as good as the best humans do. In other words, something we thought would take a few years to achieve (widespread AI replacement of human agents) will actually take many decades.
Thank goodness this is the case. Humans are critically important to making stuff happen, and will be, far into the future. Slow your roll, much money and happiness will be lost due to over estimating the speed potential of this transition, IMHO.
Totally agree, humans being critically important is the entire thesis of our work - and the "slow roll" pun is perfect in an article using convection cells as its analogy, lol!
Good catch. I just don't get why you think human inputs will be "scarce" when logic says they will be abundant. Also, why do you think capital will be abundant when logic says it will be scarce?
To clarify, not “human input”, competent human attention. The idea that is as machines become capable of offloading routine cognition as a commodity, the remaining human roles become the binding constraint. The impact of this change at scale literally shifts the macro-economic balance between labor and capital. Have a look at our first paper “Is Human Attention Still Necessary?” which outlines this argument and the potential implications carefully.
I did read it, carefully, and common sense tells me that labor will become less important in relation to productivity while capital becomes more important. In other words expenditures aimed at materials will consume more of the pie than expenditures aimed at payrolls ..... for a any given production level.
Labor will cost relatively less as it becomes LESS SCARCE (mundane cognitive work is becoming more automated) while capital will cost relatively more as it becomes MORE SCARCE (upfront costs of automation rather than extended costs of labor).
Future reality appears to be the opposite of what you envision. If your corporate model guarantees loss of company control to those that fund its foundation (critically important) and promises control to those who can be easily and inexpensively hired, your company might never get off the ground.
You differentiate between "human input" and "competent human attention" but they are one and the same. The availability of "competent humans" will increase as their jobs become automated, right?
Appreciate the back and forth, Blain, and I mean it when I say disagreement is welcome here.
One thing for the record, since it's the actual crux: the paper doesn't argue that human input becomes scarce. It separates routine cognition, which is getting commoditized, from non-routine judgment, which isn't. Your case treats those two as the same thing, and that's the reason we keep talking past each other. Freeing people from routine work doesn't make judgment abundant. It makes judgment the binding constraint. That's the whole claim, start to finish.
We see it differently and that sits fine with me. I'm going to leave it here and get back to building. Thanks for reading closely.
Judgement has always been a binding constraint, seeing things differently definitely sits fine with me. Thanks for the "discussion".