You hit the nail on the head with conveyor belts. I am a robotics engineer and I believe strongly that if done properly, automating the economy can be very good for us ("properly" to me means making sure there is some scheme for collective ownership and collective benefit, for example by encouraging cooperatives to automate), but in my mind this is best done with purpose built machines, not a bunch of very expensive humanoids.

The societal change to do this 'properly' in your terms, how can this be achieved? I've come up with the more or less same solution, but I don't see it happening.

There are incremental approaches but they would require states to promote co-operatives and there are simply no political incentives for politicians to do that.

E.g. government grants/loans, subsidies, or allowing workers to preferrentially buy out bankrupt businesses. And that's not even getting into the more radical ideas like mandating worker ownership for businesses beyond a certain size/age/revenue. Compared to the kinds of subsidies and preferential treatment that already exist for certain industries and businesses, these aren't exactly radical, they just lack a powerful lobby and donor class.

Generally I prefer Richard Wolff’s approach, which is to educate people about cooperatives and promote their development. In the same way that our current government provides all kinds of support for capitalist firms (in this use of the term, this means top down owned and managed), we should provide similar encouragement and support for the growth of cooperatives.

There are side quests like abolishing or curtailing intellectual property restrictions, which concentrate wealth in the hands of the few and make the rest of us dependent on them. (I have a whole philosophy about this but the takeaway is, no intellectual property restrictions don’t do what you think they do, and the economy and investment would work fine and actually much better without them.)

For discussion of the need for cooperatives, see any of many talks or books by Richard Wolff. This is a fine place to start:

https://youtu.be/a1WUKahMm1s

Adding to this: there are a LOT of anti-democratic laws on the books, put there by powerful people who don’t want us to have power. Legal efforts to fight those are worthwhile.

There’s also the strategy of building new economic tools that aren’t owned by anyone, via open source. That’s why I work on open source farming robots now. Building technology that can help people lift themselves up can reduce our reliance on the owners of the economy. When they put up roadblocks to our independence we protest, fight them in the courts, and build more tools to get around their control.

It’s a battle that never ends. There is no triumphant moment when everything gets solved - we need to educate, work, fight, and build. Philosophers have been writing on subjects of our collective economic well being for hundreds of years, and we haven’t achieved their dreams yet. But we can make progress one step at a time.

When you think about the manufacturing advances of the past 60 years, this is logical.

Machinists used to require skill. Now they LITERALLY just screw the part into a machine and press a green button. And the pay went from 100k+ to $45k. And the technology went from $75k to $1m. But it's still cheaper to pay a human to push a cart then it is to develop the rest of that technology.

But this is the investment that next level requires. Look at the progression (Google search this) from a Bridgeport, to a Mazak Nexus 6800, and then look at a Mazak Pallettech robot.

The first machine is completely manual. Most have retrofit digital displays and some have retrofit powered axis, but you still have to know what you're doing. The second one is semi automatic, you just need a human to set it up and an even dumber human to run it. The last machine is a robot that you bolt onto the second machine to get rid of the dumb human operator.

> it looks like an overly error prone solution looking for a problem to me

It looks like starting with a small, manageable problem before expanding the robot’s scope of work.

All of those less expensive and more reliable ways are already being used to the greatest extent possible. A modern car plant is absolutely crammed with autonomous guided vehicles and conveyor belts and any other means you could imagine of transporting stuff.

Mercedes are probably looking at final assembly for this technology, which is the most labor-intensive stage by a large margin, mainly because you've got to fit awkwardly-shaped parts into awkwardly-shaped places. A humanoid robot can potentially do useful work in this area without having to completely reimagine the final assembly line.

https://www.youtube.com/watch?v=Qlguwy9JMSA

If cheap enough, it may solve a problem for people who don't know that they need to hire engineers to make that conveyor belt, or if they don't know if the conveyor belt will be needed for more than a few months.

If it's general enough, but does simple tasks, that's still a win over 'this is a nutcracker that work only with this precise nut +-1 mm at a rate of _exactly_ 1 nut per hour +-0.1 seconds.' which may be some of the types of machines out there.

You start where you can before moving on harder tasks. A conveyor belt is never going to be able to weld curved joints and be wiring installers in the future.

I used to work for Apptronik, and for several other humanoid projects. These robots are not going to be doing tasks that are at the edge of human dexterity any time soon. They can barely drop objects into boxes.

Edit:

I saw a video a few days ago from Figure where it was putting dishes onto a dish rack at relatively high speed and that was genuinely impressive, possibly the best object manipulation I've seen. But that's still closer to dropping objects in boxes than to the welding and wiring work that I've done.

Fixed machinery already does most of the welding in automotive manufacturing. Changing that to a movable base doesn’t seem like an impossible problem especially if you can use a fixed point on the object and then use relative coordinates like a CNC machine. I could easily see a robot pulling out stamped metal parts, making a couple of welds and then stacking them up.