Robots have many advantages, for example they are able to reduce manufacturing costs. A robot at an assembly line can work longer than a human, so it makes sense to replace human work with artificial intelligence. On the other hand, robots have a big disadvantage: They are looking very similar to humans and it's possible to build robots which are more intelligent than humans.

Let us focus on the task itself. Suppose a human worker would be replaced with a robot. At first sight it's simply an automation task. The robot will increase productivity and everything is fine. But why does the robot look like a robot? If the aim is only to do some work with a machine, a normal machine without any artificial intelligence would be much better. Engineers are arguing that a highly complex task can only be automated with a human like robot. It needs the same or even higher skills than a human can provide. But are the consequences anticipated? After the work was delegated to the robot, mankind will be confronted with millions of robot workers who are more intelligent than the average person. Is that a good idea?

Update: Assumptions about robots

Robots are the result of a design process which is fulfilling the needs of the previously defined requirements. A factory robot who should be able to hold a screwdriver needs a human like hand. If mankind defines certain goals, for example that robots should increase the productivity and should handle dexterous tasks, the result is a certain shape of a robot, and a certain type of software which is able to control the machine. So we have to ask, if certain requirements are problematic.

  • $\begingroup$ I think you're asking multiple questions here: the one in the title and if, in general, it is a good idea to create an AGI or superintelligence. $\endgroup$ – nbro Jul 10 '19 at 20:40
  • 1
    $\begingroup$ Have you ever seen an assembly line? The robots putting cars together don't look anything like human beings, and are certainly not intelligent. $\endgroup$ – Oliver Mason Jul 11 '19 at 8:26
  • 2
    $\begingroup$ Can you please elaborate on the following issues? 1. Why do robots have to look very similar to humans? 2. Where is a robot that is more intelligent than a human? 3. Which engineers are arguing that we need human-like robots? As it is, your question makes rather a lot of assumptions that I don't think are true. $\endgroup$ – Oliver Mason Jul 11 '19 at 8:55

We aren't putting general intelligence on the assembly line

The machines you're talking about (those who could be more 'intelligent' than a human) are both a long way off and over kill for an assembly line.

Any AI on an assembly line may be brilliantly suited to deciding where and when to perform tasks, which products aren't up to standard and so on...but show it a chess board and it wouldn't understand it in any terms other than those of an assembly line. Likewise an AI that can beat you at chess wouldn't be able to solve problems on an assembly line or drive a car...

| improve this answer | |
  • $\begingroup$ I would even go so far and say that assembly line robots are not making use of AI at all, because they don't have to. Some applications (eg sorting items) might make use of pattern recognition, but welding car parts together is not an AI task. $\endgroup$ – Oliver Mason Jul 11 '19 at 13:30
  • 1
    $\begingroup$ @OliverMason I agree but pattern recognition for sorting or some machine learning model for quality assurance are still called AI - I expect some news outlets may well report some headline like "Artificial intelligence to replace human workers in car manufacturing" and where it actually means pattern recognition people are picturing Wall-E with a welding torch. Rather than refute the AI is used I thought it best to try to point out the difference between "AI" and a general intelligence. $\endgroup$ – Lio Elbammalf Jul 11 '19 at 14:16
  • 1
    $\begingroup$ I like your point about overkill--it would be ridiculously expensive, computationally, to use a hypothetical AGI for repetitive tasks. $\endgroup$ – DukeZhou Jul 11 '19 at 16:45

Never forget that intelligence is a spectrum, and a relative measure of utility vs. other intelligences, in the context of a task/problem or set of tasks/problems.

So, while it's possible to create a robot that performs better than humans in repetitive tasks, such as an assembly line, creating a robot that is better than the average human at all tasks engaged by humans is a tall order, and not possible with current tech.

As to why one might make a robot that looks human, the motive may be to increase the (still limited) set of tasks the robot can perform. But it seems inefficient in terms of what we use robots for today. (There is a new police robot in the field, and it's shaped like a bullet, not a human.)

"A factory robot who should be able to hold a screwdriver needs a human like hand." This is a misleading in that the human hand is not the only structure that can grip and turn a screwdriver.

Many jobs will be lost due to strong-narrow AI and automation, but AGI (robots as intelligent as an average human in all tasks engaged in by humans) is still science-fiction!

| improve this answer | |
  • $\begingroup$ The promise, that technical difficulties will prevent Artificial Intelligence was first given during the Lighthill report in the 1970s. The same forecast was given during the last AI winter in the 1990s and now the argument is given, that AGI is technically not possible. How long will it take, until science will proof the opposite? $\endgroup$ – Manuel Rodriguez Jul 10 '19 at 21:26
  • $\begingroup$ @ManuelRodriguez at present, AGI is in the realm of fusion power, always off in some future. I do believe it's possible, but not with current processing power, memory, and methods. (Physics has a way of rearing it's ugly head;) Computational complexity theory helps explain why it's such a daunting task. $\endgroup$ – DukeZhou Jul 10 '19 at 21:40
  • $\begingroup$ The current "AI revolution" itself seems to be a function of contemporary processing power and memory, in that the concept of Neural Networks has been around since the 1940's. And even the most powerful contemporary AIs, although legitimately strong, are extremely narrow in terms of scope of capability. $\endgroup$ – DukeZhou Jul 10 '19 at 21:43

Even in the easiest tasks handled by robot the robot can manage more efficiency than humans, the issue is that only on narrow field. Some more complex tasks can be handled too, if we combine computing and automation and many individual robots. Many factories work nowadays with minimal human intervention and there the efficiency is sort of over human.

Narrow AI is over human on games. So, we are already in a situation where robots / machines / you name it make things more efficiently than humans. So, I would observe around and see what is currently happening. Some job descriptions disappear and new ones arise. No chaos so far.

We have the chess player that plays chess with itself as opponents millions of rounds and which made alien super human moves, when game was learned. To call that AGI, or to be afraid It becomes one, we are far away, it knows nothing about surgery, delivering post nor talking. Easy tasks for intelligent robot with AGI in brains, no case to that machine.

So about the title:

Does it make sense to invent intelligent robots, if we only need to automate the economy?

You answered that already yourself:

Some tasks will need It.

So, if we want maximal outcome is achieved by something over human effort and we want to achieve it, so why not?

This was tagged with philosophy so some deep thoughs in the end.

First of all,

An artificial intelligence system being developed at Facebook has created its own language. It developed a system of code words to make communication more efficient. Researchers shut the system down when they realized the AI was no longer using English.


An early skynet? Not that but a risk, yes. As every machine, also AI may perform errors. The risk is not (yet?) the amount of intelligence or self awareness but the thing that AI is no more trackable and thus controllable in this sense.

Also ethical desicions are always mentioned at this point: should they be given to machine and who is responsible if machine fails. To this philopsophical aspect I cannot say much, although this maybe interests you most. In short I can say, that humans make the machines and you cannot sue only the machine.

To the very question of need for development, well we are too deep already.

| improve this answer | |
  • $\begingroup$ The answers mentioned some examples in which Artificial Intelligence is stronger than humans today, which is computer chess and a chatbot from Facebook. It was also explained, that mankind has a need for AGI and a higher efficiency. This kind of technocracy is the opposite of an easy to manage future and implies that mankind has lost control over his invention. $\endgroup$ – Manuel Rodriguez Jul 12 '19 at 17:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.