One needs to recall that manufacturing labor redistribuion in reponse to import liberalization is supposed ot happen in part toward export industreis. Two thing stood in the way of this. a) US trade negotiators priortized opening foreign countries to US service and intellectual property exports, not manufactured exports. b) seculare defcits produced trade deficits instead of balance of export surplusses.
The instinct to compare who is impacted, what their ability to adapt is and how that's localized is a good one. I think that Eloundou is insufficient alone for the purpose of modelling. It was a good early effort, but analysis should be incorporating more than this and modelling more than static effects per task. One approach is adding elasticity based on historical data.
I won't claim to have this all figured out, but I would say that anyone taking efforts should apply something at least one level of depth more sophisticated than Eloundou before the other analysis.
The cause of Americans' job losses, whether to outsourcing, automation or AI is obvious and we've had since 1990 to address this, but refuse to do so. When Americans' wages are significantly high, employers look for alternate sources of labor to save money. The US has 5% of the world's population, yet consumes 25% of its resources. Does that level of consumption mean Americans live better than people in other countries? Many middle income countries such as Mexico has a high Human Development Index of 0.8 vs the US very high HDI of 0.93, yet Mexico's per capita GDP is 1/7 of the US. Clearly there is a low return on the additional income of Americans and we are not living as efficiently. American society is built for cars. Cars not only cost $12,000 a year to own (the same amount as the per capita GDP of Mexico), it is only in use 1 hour a day and sits unused 96% of the time. Even worse, cars enable sprawl. +50% of American cities' land is used as infrastructure for the automobile. A US single family home's lot only uses 20% of its space for living. The remaining 80% consist of garage, driveway, front yard and backyard. This wasteful use of land does not scale well and when we run out of space to build, housing prices skyrocket. In contrast, Mexico did not have the wealth to build its society for cars, but for people. Mexicans homes use 80% of their lot space for living while the remaining 20% in the middle of a home can be used as a patio, a parking space for an automobile or Frontage for a home business. Rents are around $400 because housing is plentiful. Mexicans' neighborhoods are compact, mixed use, walkable, and there is cheap and plentiful public transportation. Americans' two biggest costs are not problems for Mexico and many other countries that are not car dependent.
One way to get away from car dependency is for localities to convert parking lots of big workplaces to dense affordable, mixed-use rentals for those that work nearby. These 5-minute communities can get rid of 250 hours per year of commute time, the car, garage, driveway, front and backyard in favor of a common pool of shared vehicles and common greenspaces. The savings from efficient use of space and elimination of personal automobiles alone could save workers half their expenses. There are many more opportunities for sharing to cut costs, live sustainably, improve quality of life and lessen pressure on job losses in these kinds of communities.
Fascinating! I keep hearing all these generalizations about the impact of AI, and it's a lot more complex than people realize.
One needs to recall that manufacturing labor redistribuion in reponse to import liberalization is supposed ot happen in part toward export industreis. Two thing stood in the way of this. a) US trade negotiators priortized opening foreign countries to US service and intellectual property exports, not manufactured exports. b) seculare defcits produced trade deficits instead of balance of export surplusses.
The instinct to compare who is impacted, what their ability to adapt is and how that's localized is a good one. I think that Eloundou is insufficient alone for the purpose of modelling. It was a good early effort, but analysis should be incorporating more than this and modelling more than static effects per task. One approach is adding elasticity based on historical data.
My own approach is categorization into unbounded, bounded and adversarial tasks. This model brings in Jevons (adversarial) and Baumol (unbounded) effects similar to elasticity effects: https://substack.norabble.com/p/ai-jobs-the-hidden-rules-of-demand
I've been slowly working on modeling to match the conceptual model: https://github.com/norabble/ai-labor-exposure
I won't claim to have this all figured out, but I would say that anyone taking efforts should apply something at least one level of depth more sophisticated than Eloundou before the other analysis.
The cause of Americans' job losses, whether to outsourcing, automation or AI is obvious and we've had since 1990 to address this, but refuse to do so. When Americans' wages are significantly high, employers look for alternate sources of labor to save money. The US has 5% of the world's population, yet consumes 25% of its resources. Does that level of consumption mean Americans live better than people in other countries? Many middle income countries such as Mexico has a high Human Development Index of 0.8 vs the US very high HDI of 0.93, yet Mexico's per capita GDP is 1/7 of the US. Clearly there is a low return on the additional income of Americans and we are not living as efficiently. American society is built for cars. Cars not only cost $12,000 a year to own (the same amount as the per capita GDP of Mexico), it is only in use 1 hour a day and sits unused 96% of the time. Even worse, cars enable sprawl. +50% of American cities' land is used as infrastructure for the automobile. A US single family home's lot only uses 20% of its space for living. The remaining 80% consist of garage, driveway, front yard and backyard. This wasteful use of land does not scale well and when we run out of space to build, housing prices skyrocket. In contrast, Mexico did not have the wealth to build its society for cars, but for people. Mexicans homes use 80% of their lot space for living while the remaining 20% in the middle of a home can be used as a patio, a parking space for an automobile or Frontage for a home business. Rents are around $400 because housing is plentiful. Mexicans' neighborhoods are compact, mixed use, walkable, and there is cheap and plentiful public transportation. Americans' two biggest costs are not problems for Mexico and many other countries that are not car dependent.
One way to get away from car dependency is for localities to convert parking lots of big workplaces to dense affordable, mixed-use rentals for those that work nearby. These 5-minute communities can get rid of 250 hours per year of commute time, the car, garage, driveway, front and backyard in favor of a common pool of shared vehicles and common greenspaces. The savings from efficient use of space and elimination of personal automobiles alone could save workers half their expenses. There are many more opportunities for sharing to cut costs, live sustainably, improve quality of life and lessen pressure on job losses in these kinds of communities.