It’s the same thing. Anything that could be automated but isn’t automated is because it isn’t cost effective to do so or there is insufficient capital to invest. These are resource allocation issues. You can’t just wave that away.
No it isn't. Using a script to automate a process frees me from having to carry out the process manually. It has nothing to do with central economic planning.
Most things in the world require physical processes. Automating those is quite a different task, and requires resources. How would you go about automating sewing a shirt? How about picking strawberries?
Again this is tech hubris and a lack of understanding of economics and history.
> Most things in the world require physical processes.
No one ever disputed that. The principle still holds if we apply this logic to physical processes; by automating or reducing the labor necessary to conduct a physical process, I can enjoy the benefits of the process without having to engage in the labor of the process.
> How would you go about automating sewing a shirt?
> Anything that could be automated but isn’t automated is because it isn’t cost effective to do so or there is insufficient capital to invest. These are resource allocation issues. You can’t just wave that away.
This is true in the long run and I suppose the argument you are making is that any attempt to interfere with the present system of resource allocation will constitute a centralization that would be less effective than free market capitalism, so the notion that we could redistribute the surpluses generated by labor saving devices to the average person is inherently a call to economic centralization. This might be true, but I would propose an alternative reading:
The surplus of labor-saving devices has primarily accrued to the owners of these devices. You might then claim that these owners are owners because they have found a means of servicing a market demand. Each dollar they possess is a vote from the market that these guys really know what they are doing, and that the world wants more of it. If we were talking about spherical billionaires in a vacuum, I'd agree with you - but this issue is complicated by the compounding impacts of inheritance and its correlation with access to credit, as well as with the existence of competitive moats (e.g. network effects, intellectual property, sunk costs, natural monopolies, etc).
The optimistic read of the technology sector in the 2010s was that businesses would compete with one another to provide services that would ultimately improve people's lives. Instead, we got Windows 11. That wasn't a consequence of users voting with their dollars, it was a consequence of Microsoft entrenching itself into workflows that cannot be economically altered in the immediate future. There are lots of examples of the market not being particularly effective at economic allocation if we step outside of the logic that any purchase is a revealed preference which indicates approval of the good or service being purchased. Apply this logic to the purchases of gamblers, alcoholics, drug addicts, or murder-for-hire plots and the limitations of the logic become obvious.
No my argument is that trying at a broad systemic level to make specific outcomes happen is susceptible to the information problem. Trying like the op suggested to automate away work is utopian and improbable at best. If you squeeze labor out of one kind of drudgery you have no way of predicting the results, and you’re certainly not going to end up with Star Trek.
To my minor aside, look at that shirt. You have to essentially glue the fabric into a board and all the robot can do is a rudimentary set of side seams and sleeves on the tshirt. There’s no finishing work on the collar or hem so it’s useless. That robot exists as a demo and is used precisely nowhere. You could in theory do this, but it makes no sense economically.
And yes I’m aware of Japanese strawberry picking robots. You’ve clearly misunderstood what I’m saying. These thing may be technically possible but they remain infeasible for other reasons.
> No my argument is that trying at a broad systemic level to make specific outcomes happen is susceptible to the information problem.
This is exactly what I said you would say:
> I suppose the argument you are making is that any attempt to interfere with the present system of resource allocation will constitute a centralization that would be less effective than free market capitalism
Further:
> Trying like the op suggested to automate away work is utopian and improbable at best.
We are a long way off from the self-replicating systems that could feasibly make work effectively optional, but you haven't made a convincing argument as to why it is improbable that automation could reach that point.
> And yes I’m aware of Japanese strawberry picking robots.
You clearly were not aware of them or you would have picked better examples. Your original comment consisted solely of the statement: "It's the same thing." and now you're continuing with that flippant attitude by pretending that I'm misunderstanding your argument when I anticipated it in its entirety.
I clearly was aware of them. Do you think I just rattled of the bit about needing special glue to hold the fabric and only certain seam types being possible? There was a whole thing about these in the Economist last year and it was discussed on HN. While it’s technically possible you can’t deploy it. It turns out gluing an then applying solvents to fabrics doesn’t result in a product people want.
This Star Trek stuff is improbable because everything has to be coordinated somehow and waving your hand and saying magical future ai is the only proposal anyone ever has. So yeah, maybe super advanced AGI could do it, but probably not. We don’t even have good models now of how large economies work down to a granular level. People are like I said messy and respond in weird ways to their environments. The best we can do right now is working with prices as signals for the amount of effort other people are willing to put into something. And while that’s imperfect, it’s just improbable that we can do much better. Which is not to say that narrow objectives aren’t possible, only that the bigger and broader you aim the more impossible it becomes.
You cited them as examples of tasks that would be difficult to automate. The pickers have been commercially deployed for the last four years.
> This Star Trek stuff is improbable because everything has to be coordinated somehow and waving your hand and saying magical future ai is the only proposal anyone ever has.
Redistribution already occurs without the use of an AI.
> You cited them as examples of tasks that would be difficult to automate.
Yes because they are. I specifically gave an example where a machine exists but it's impossible to use for the real world, and an example where economics generally prevent adoption. That gets to my whole point.
> The pickers have been commercially deployed for the last four years.
Yes narrowly, and in only a few places where there are extreme labor shortages.
You are clearly misunderstanding me.
> Redistribution already occurs without the use of an AI.
I didn't make the claim that it didn't happen.
I feel like you're willfully ignoring what I'm saying. These things are hard and rolling them out universally often doesn't work because it is either impractical or economically infeasible to automate things or you run up against regulatory/cultural/material issues. The best we can do is piecemeal progress where incentives align.
Hah, I wish that were the case. A whole lot more things would be automated if that were true.
Automation requires resources, but it also requires vision, cooperation among affected parties, a workable regulatory framework, maturity and availability of required solutions, and availability of competent integrators. There are all kinds of reasons something remains manual besides mere resource availability. And all those things change over time.
There's not much you can do about most of those things, but becoming a programmer and working to develop better solutions is one way to make a difference. Even if you don't work directly in automation, your work can trickle down to the people like me who do cencern themselves with automated sewing and strawberry harvesting.
What I mean by resources is the things you mentioned inclusive of vision.
I picked those two examples because you can literally build a robot to do it, but it is either unworkable in the case of the shirt or financially not viable like the strawberry robot.
Using your model, no technological development would ever occur because the fact that something had not happened yet would indicate that it could not possibly happen due to a lack of resources. This is the anecdote about two economists walking down the street and refusing to pick up a $100 because everyone knows that in an efficient market, someone would have already picked it up.
At some point the resources necessary for development are there but the technology itself has not actuated. This invalidates your original claim that: "Anything that could be automated but isn’t automated is because it isn’t cost effective to do so or there is insufficient capital to invest."
It should be obvious. There are plenty of thing we can build robots to do, but we don't because its wildly more expensive than we can sell the resulting product for. We can mostly automate construction but it turns out the land acquisition dominates costs, installation still ends up being sloppy and human, building codes are different everywhere, and people want a different kind of dwelling than what prefab is suited for at the moment.
If it should be obvious then the evidence should be equally obvious.
Or perhaps the world is a bit more nuanced and it may very well be that we're stuck in some local maximums that our current methodologies don't allow us to escape but escaping them is relatively easy if we chose to implement a meagre amount of resources for that purpose which is something we don't do because we're stuck in that local maximum and so on and so forth.
Another way of looking at what you're saying is that we're doing things optimally and that there's no room for improvement when that very obviously is not the case.
There are many gross inefficiencies in our system as it currently is -- look at food production for example. How much of the food produced globally is outright wasted? 30%? 50%?
If we made a conscious effort to tighten that up we could reallocate those resources to solving the problems of automation issues that you're describing.
The true hole in one for automation is a durable machine that can make a copy of itself as well as useful economic goods. Bonus points if this machine can be in a humanoid form to integrate into our existing economic infrastructure.
Once you have a self replicator you can have it make as many copies as needed to solve any problem you need with minimal human effort.
But a self-replicating machine isn't on anyones radar. Have you ever seen a politician or policy person discuss this?
I’m sorry, any talk about self replicating machines is just science fiction at this point. It’s not a serious thing to discuss as a near future possibility.
