Hacker News Reader

The thing about automation as it regards to economics is that it doesn't have to be anywhere near a complete solution to have a huge impact. You don't need factories full of fully self-sustained robots to cause massive underemployment.

If you can partially automate a task to the point where 1 person can do the work that 2 people used to do, you've potentially eliminated 50% of that job's market. People are somewhat adaptable, so some amount of this sort of change can be absorbed (and has historically been absorbed) as people retrain for other types of jobs, but we're quickly approaching a point where the changes brought upon by automation are going to outpace people's ability to adapt to the changes, IMO.

Absolutely true. I think when people say that robots are going to replace low-skilled workers only the uninformed are imagining an entirely self-sufficient factory that only has one door that the raw materials goes into, and another door that the product comes out of.

What we really should be imagining is a factory floor where instead of 50 people on an assembly line, or instead of 20 waitstaff at a restaurant, you only need one or two guys who know how to replace parts on the robots

If it takes 50 automation engineers to figure out how to automate a factory... once they figure this out, it doesn't need to be figured out again. (Not a big problem for them, there's always another problem to solve.)

But not only is the entire factory out of work... all factories doing the same task are.

If someone perfects a shelf-stocker, it doesn't matter that they needed the help of 1500 engineers... at that point, all grocery stores need far fewer employees. So does Walmart and Target. Even those crappy little stores hanging on by a thread... if they can scrape together the $50,000 for a robot, that's what they'll do.

These things only have to be figured out once.

It'd take 50 automation engineers a year to figure out how to automate a single step in a factory... just in time for the product redesign that may make the step irrelevant.

These things may have to be redesigned from scratch every time a product is updated. iPhones don't have all of their screws, screens or cables in the same spot from revision-to-revision.

You can only automate product development when things barely change.

> It'd take 50 automation engineers a year to figure out how to automate a single step in a factory..

Higher levels of abstraction make the process easier. What was nearly impossible to automate 10 years ago has become nearly commonplace today. The pace of this abstraction is only increasing as the availability of robotics increases.

I doubt that but like everything else better tools will make this process more efficient...

I've seen SMD pick-and-place robots at work that would make you believe you're living in a science fiction novel.

http://www.youtube.com/watch?v=SRu02F6AOmg

For example. There are faster ones but with this one you can still follow the action. (est. about 8000 cph, 20K cph is available right now and is too fast to follow directly, you'd need to videotape it and then slow down the movie, it places 6 parts per second, when you look at one up close it looks like parts simply appear on the board out of nowhere). Not all that long ago that work would have been done by a small army of people, and that board would have been 10 times the size. Automation works iteratively, after enough steps it becomes worth it to design to the tooling rather than the other way around. So at some point your phone won't have screws or cables, and the screen will be sandwiched right on top of the board if that's the most economical. Cables are a source of unreliability anyway and screws are made for dis-assembly whereas the trend is to concentrate on the assembly phase with dis-assembly (and repair) becoming ever more difficult.

Reballing BGA's is so much fun.

PCB creation has been extremely automated for a long time now. Between pick-and-place (around since the 1980s) and reflow soldering (auto-magic soldering), its safe to say that PCBs have been automated for the last 30 years at very least.

So yes, its impressive that modern pick-and-place is so fast, but really, it was a solved problem that has only become "more solved" as the years have passed.

In contrast, actual _assembly_ work is still very human based. The iPhone is famous for this. There are exceptions of course (the Sony Walkman was famous for being a purely automatic design from start to finish... the Tesla Car is mostly automated IIRC as well). But the typical products of the modern person (iPhones and Ford Cars) are still largely assembled in human assembly lines.

I think you may be discounting non-linearities in the progress of automation due to machine learning and AI; assuming AI makes massive strides in 100 years (which seems almost certain), there may be very very little (if anything) that humans can do better (or more cost effectively) than intelligent machines.

In virtual space that is more meaningful than in physical space. Manufacturing automation relies on fixtures, tooling, and material movement. Having been a industrial engineer in a variety of industries, I can think of only a small handful of products that would lend themselves to full automation, and even those would require enormous upfront capital costs.

I appreciate your point, but I think you're wrong, given a sufficiently powerful AI -- one that is truly more intelligent than a human, which is a real possibility in the next 100 years.

The smarter-than-human AI would be better than us at everything, including designing, controlling, repairing, programming, etc., a general-purpose humanoid robot

I appreciate the realism you bring forth.

Nonetheless, I don't think people are talking about lights-out levels of automation (which exists, but is horribly complex for reasons you've mentioned).

Realistically, we're looking at robots that are going to take over a ton of jobs. The new-fangled example is the Fully Automated Cow Milking Machine, which will be the bane of milkers everywhere within the next few years.

http://www.lely.com/en/milking/robotic-milkingsystem/astrona...

Sure, its a $200,000+ piece of equipment, but costs are going to drop and it will eventually become commonplace in farms.

Similarly, lawyers have already lost a ton of money to automation. Creating a company no longer requires consulting with a lawyer... instead you can purchase a 1-2-3 piece set from a website.

http://www.legalzoom.com/additional-business-services/s-corp...

First of all: you don't have to automate the entire workforce in order to have devastating unemployment, inequality and poverty. Every automated job means one fewer job on the market and one more unemployed person competing with everyone else for a job. You don't need self-diagnosing and self-repairing machines to wipe out entire classes of jobs. Retail, fast food and delivery/freight are huge examples of job categories that are coming perilously close to total automation.

Secondly: automation, like blackhat hackers, never gets less sophisticated. When a problem is solved by automation it stays solved. The sum total of the capability of all automation in the world increases monotonically over time. Human capability, on the other hand, tends not to increase at all.

From what I've heard, automation is actually, really on-the-way. I've forgotten the exact conversation and this is certainly hearsay, but a friend of mine was describing a conversation with a upper level person (VP, CxO or board member) and was hearing about how 2014 was the year Starbucks would start to get rid of the "barristas". [Yeah, they used to be icons of Starbucks...]

>then have some other machine to fix all those things automatically >on the fly. Building self-diagnosing and self-repairing machines is very complex.

So only do the first part of the automation: automate the garbage collection truck and have the truck pull over when it's sick... That captures 95% of the value; we can leave self-repairing for later.

>So, where is my flying car?

1) waiting on a suitable power source; 2) waiting on automated driving. It'll probably still be waiting on (1) and not (2) in 10 years.

here is your not-so-cheap flying car that is probably not available yet: http://www.gizmag.com/flying-jet-car/32287/