Will Robots replace Humans?

If you discard the debate about artificial intelligence for a moment and assume that at some point we did manage to create sentient robots there is a question people ask - Will robots replace humans? - My answer to that is no, but that requires some explanation.

First of all you need to define the word "replace" and what exactly you mean by it.  In my definition for robots to replace humans they would have to take our place and do what we do.  In this respect you can define humanity by a number of characteristics, notably survival instinct, and the instinct to procreate.  The latter I realise some people will take issue with who have no intention of having children, and as a gay man I would agree that this does not define my life personally but as a race as a whole in order for humanity to survive it needs not only to continue living but to continue growing in number, or at the very least maintain a steady population which both require procreation.

As we have defined two core characteristics of humanity you can apply those to robots in the scenario we are contemplating and ask would they act in the same way and to me in my opinion they would not.  First of all survival is something that is very hard to program into a machine quite simply because survival is something for a machine that is very hard to define.  If we define it as the continuation of operation, many machines can continue to be functional while maintaining low power and little or no activity - in other words they can exist in standby mode.  If a machine entered standby mode with no intention of ever leaving it, would you consider it as surviving?  It's still powered and still operational, it's just not doing anything.  To borrow a human analogy if you were to voluntarily enter into an indefinite coma with the intention of staying in it, would you consider yourself as surviving?  What if you stay in it for 10,000 years and then die by freak accident?

In terms of procreation for machines that is closely linked to the above, for a machine there is little to gain and no real motivation.  As humans we will eventually die we therefore need to be replaced.  If humans lived potentially forever, there would be less motivation to reproduce.  Add in logical conclusions a robot would come to in response to limited resources and non-renewable energy constraints versus their energy demand, even with increased efficiency it would be unrealistic to say they would not be limited as humans are in terms of sustainable growth.  Relying on logic alone a robot would conclude that reproduction beyond an optimal population would be unnecessary.

Given that I therefore think that Robots would not take up our characteristics they would therefore not "replace" us.  Of course this is all from a human perspective.  This is what survival can be defined as from our behaviour, there's no way of knowing what survival would be defined as to a robot.  As for the argument that robots will use the definitions we provide, that relies on classical artificial intelligence - we use that less and less because it is restrictive and does not advance that far.  Modern artificial intelligence uses the evolutionary algorithm, which allows the agent to evolve as they see fit to achieve their task, including rewriting their own codebase, as a result the agents which evolve take complex forms of machine intelligence that are beyond our understanding and beyond our control as they are not explicitly designed by the programmer.  Any intelligence we successfully create capable of sentience will inherently be therefore beyond our control.