Hacker News Reader

Comment by eatitraw | Hacker News Reader

Can someone point to me why everyone is so sure there will be robots?

Let me explain why I am asking. It seems to me that we already hit some sort of physical limit for the clock rates. We can't (easily) double clock rate of a CPU, so we add more CPUs. But programming multithreaded applications is more difficult, so it is more difficult to utilize this power.

Can't we hit similar limit for humans? Is there a limit on what humans can comprehend and build? Let's say one programmer can code 1 abstract feature-unit per day. There are limits. What makes everyone think that replacing humans with robots will require not 10^100 abstract units of features? Why no one considers that this problem may be just _too difficult_.

View on HN

Partly because the robots we're talking about here don't necessarily have to be of the humanoid/strong AI type.

Bear in mind that rug weaving was considered too complex and intricate to be automated, until it was. Also bear in mind that humans used to build cars in factories, now robots do, and that the very term "computer" used to refer to something a human being did, rather than a machine.

The complexity of the necessary systems don't need to appear all at once, but can be built on top of, and feed off of, one another. You replace jobs piecemeal -- first the simple labor-intensive jobs, then complex manufacturing, then automate logistics. At some point, autonomous cars are delivering goods manufactured entirely by robots on behalf of autonomous corporations.

Eventually, as technology and AI improves, you could have creative works done entirely by machines, or almost entirely. Movies written, directed by, starring, composed, engineered and distributed entirely by AI, with books and music and toys and everything.

There's no evidence that it's too difficult, but plenty of evidence of things once considered impossible, which turn out now to be commonplace, or taken for granted.

There are many dimensions to it, but it boils down to the fact that for many people worship of Technology has replaced worship of more traditional Gods.

That being said, there is nothing inherently impossible to prevent robots from being deployed in large quantities. Not the kind of robots you see in SF artwork - capable of displaying human-like levels of agency, creativity, wit, etc - but dumb little machines that can repeat the same dumb little task once, and twice, and a million times without a single error. That has been done many times in the past with varying degrees of success. The problem is that is the increase in speed/accuracy of little task does not produce an economic return large enough to justify the automation, the automation will never get implemented.

As a matter of fact, if the total economic return of the little task is not large enough, the task will not get done either manually or automatically. Paul Graham has already written about how changes in perceived social value of a given line of work has rendered it obsolete with no need for technological disruption whatsoever.

So there are several thresholds in the set of all economic activities at any given time. Below the outermost threshold the task will be procrastinated away until the whole environment changes and makes it irrelevant. A little above that there is another threshold that makes a task impossible to make a living out off, but not unimportant enough to be ignored for ever... so you have a lot of unpaid labor done on the sides to take care of that (think domestic chores). Then there is the formal economy with several layers of skilled and unskilled labor.

At each layer, robots can eat a chunk of that... depending on the technical and economic feasibility. By example, we have Roomba robots to clean the floors because the task is relatively easy and can be solved for a price people is willing to pay. On the other hand, we do not have (yet) robotic spiders that pick up, wipe and put back every random trinket in your living room without breaking anything because the task is too complex (but might turn out to be solvable in the future). And we do not have refrigerator-stove hybrids with robotic arms that pull ingredients out of cold storage and cook homemade meals because even though the problem is relatively well defined the price would be so great that few customers would be able to afford those, and it will probably continue to be so regardless of advances in technology.

There was a research that has shown that in places in cs, software innovation contributed to far greater improvements than moore's law. There's no reason why software improvements should stop.

Reply on news.ycombinator.com