Machines

[Antoninus Verus]

If we were to take an island and put on it a totally self sustaining factory that could produce machines, do you think the machines would eventually, given enough time, the machines would grow or evolve?

The factory would be able to mine raw ore, process it into machine parts, and construct new machines with no human intervention or mantinence. The machines would be programmed for self presevation. The factory would make robots for mantinence, mining, assembly work etc etc.

Now the question is, would a robot..."colony" like that, ever evolve or grow? Would the machine realize it could make better versions of the robots it already made to increase the chances of survival of itself?

 

[HouseApe]

Now the question is, would a robot..."colony" like that, ever evolve or grow? Would the machine realize it could make better versions of the robots it already made to increase the chances of survival of itself?

You would have to program the machine to introduce variation in the robots. If the variation causes the robot to improve in some way, make all new robots have that variation, and selectively kill off the old robots. So, my answer is yes.

 

[Antoninus Verus]

You would have to program the machine to introduce variation in the robots. If the variation causes the robot to improve in some way, make all new robots have that variation, and selectively kill off the old robots. So, my answer is yes.

I mean could a machine do that without being programmed to do so?

IE: The machine is programmed to survive in the most efficent way possible. The ore hauling robots use small wheels to move, the machine calculates that the robots would be more efficent with larger wheels so it upgrades the wheels to a bigger size in order to allow itself to survive more efficently. It would take the old modles of robots and phase them out or simply upgrade the existing models.

 

[TheMagi]

I mean could a machine do that without being programmed to do so?

IE: The machine is programmed to survive in the most efficent way possible. The ore hauling robots use small wheels to move, the machine calculates that the robots would be more efficent with larger wheels so it upgrades the wheels to a bigger size in order to allow itself to survive more efficently. It would take the old modles of robots and phase them out or simply upgrade the existing models.

Of course, it is theoretically possible. It has been done, after all, with computer code, and can be quite an efficient way of operating.

Practically, I doubt it. Living things can evolve because they are extremely versatile, capable of adapting to new biological niches, and operating effectively with quite major changes. Also, they operate not according to a close set program but according to needs which can ne met in a number of different ways.

Unless carefully programmed to adapt (which isn't the same thing as evolving) I would say that human-designed machines as we know them aren't generally capable of this.
Perhaps more importantly, there would be no reason to evolve. The machine already works. In the fossil record, most major evolutionary changes occur after major extinctions, when things don't really work - i.e necessity os the mother of evolution as well as of change.
Magi

 

[Antoninus Verus]

Of course, it is theoretically possible. It has been done, after all, with computer code, and can be quite an efficient way of operating.

Practically, I doubt it. Living things can evolve because they are extremely versatile, capable of adapting to new biological niches, and operating effectively with quite major changes. Also, they operate not according to a close set program but according to needs which can ne met in a number of different ways.

Unless carefully programmed to adapt (which isn't the same thing as evolving) I would say that human-designed machines as we know them aren't generally capable of this.
Perhaps more importantly, there would be no reason to evolve. The machine already works. In the fossil record, most major evolutionary changes occur after major extinctions, when things don't really work - i.e necessity os the mother of evolution as well as of change.
Magi

The machine would modify itself in order to survive better. It would be programmed to survive. And in order to more efficently do that, it would be allowed to change itself and its own programming so long as the core remained intact. The reason would be to increase chances of survival of the machine. The machine would basically told to give itself the best possible chance of surviving

 

[holyorders]

It can't be done.

It is a Matrix-type fantasy



The logic is simple- finite beings are limited and, at best, could only create robots of equal power. Robots have the ability to out-perform man but they can only grow and adapt at man's pace.

 

[HouseApe]

The machine would modify itself in order to survive better. It would be programmed to survive. And in order to more efficently do that, it would be allowed to change itself and its own programming so long as the core remained intact. The reason would be to increase chances of survival of the machine. The machine would basically told to give itself the best possible chance of surviving

I think it could be done. Artificial intelligence is possible to a degree. The key thing is to program the machine to maintain a high degree of diversity at the expense of efficiency. That way the possibility of survival of some robots after a catastrophe is increased. If the catastrophe causes the big wheeled robots to stop working but the small wheeled ones continue, then the machine might not die.

Much like modern ecology. We humans are efficient at reproducing ourselves and exploiting the environment at the expense of the overall diversity of life. In the event of a catastrophe, we might all die, and take the rest of life with us.

 

[HRE]

Theoretically, yes, but robots generally code perfectly.

DNA, on the other hand, doesn't. It mutates.

 

[Antoninus Verus]

Theoretically, yes, but robots generally code perfectly.

DNA, on the other hand, doesn't. It mutates.

The machine would compensate for that because it would be programmed to find the most efficent way of ensuring its survival and make changes to the robots around it and itself.

I think it could be done. Artificial intelligence is possible to a degree.

You wouldnt even need AI. The computer would change the robots that sustained it and itself in order to fulfill the programming directive of efficent self preservation. A wheeled vehicle doesnt do very well in deep sand, but a tracked vehicle moves better, so if the machine need something to go across deep sand, it would create a new variant of an already existing model to increase efficency

 

[HRE]

The machine would compensate for that because it would be programmed to find the most efficent way of ensuring its survival and make changes to the robots around it and itself.

Then I should note before this analogy goes any further that this is artificial selection, and not in any way comparable to natural selection and evolution.

But go on...

 

[Antoninus Verus]

Then I should note before this analogy goes any further that this is artificial selection, and not in any way comparable to natural selection and evolution.

But go on...

But the process, the machine would evolve. Evolution is the process of improving life and the chances of survival through physical changes

 

[David Gould]

But the process, the machine would evolve. Evolution is the process of improving life and the chances of survival through physical changes

Yes, but evolution works through natural selection. Things can physically alter without evolving - if I glue a mechanical arm to my back that is a physical change but it is not evolution.


Further, how would the computer be able to predict what would work better? In other words, how would it know what would help the machines survive better?

 

[James T]

The experiment is already under way. We are already creating our replacements.

 

[kedaman]

Its been said that Artificial intelligence is an oxymoron, machines are by definition things that were created by man, and can thus not think for themselves but do what we program them to do. That something can create something better than oneself is another contradiction, for a thing is as good as all it can do. If I create a computer and let it calculate pi to fourbillion decimals, then I can compute pi to fourbillion decimals.

 

[James T]

I hope you find that thought comforting. The Church-Turing thesis is unprovable (as with most physical laws) but states anything computable can be done by a Turing machine. Simulations of the brain are computable, in theory, my own view is that massive simplification to lose the biochemical supports elements and focus of the key functions will be possible making the computational task much simpler. I'll accept that the code is still complicated.

However, thinking programs do exist, they are very smart in narrow fields and many of the tools required once the core motivating code requirements are sorted will almost be available shrink wrap off the shelf (voice recog, image recog, math, logic etc.). Project the changes in the last 100 years to the next ... and don't be so confident.

 

[kedaman]

In a finite amout of steps, with finite amount of states, with a finite set of instructions, with a finite amount of symbols on an infinite tape.. for the moment everything beyond that we cannot construct with our finite minds. If you want to argue that we can simulate brains, I'd like to quote Skinner: The real question is not whether machines think, but whether men do. Programs do not think, that's an oxymoron. A program is a function; an algorithm, if it goes beyond that then its not a program.

 

[TheMagi]

oxymoron. A program is a function; an algorithm, if it goes beyond that then its not a program.

What if an algorithim can do what it cannot predict, except by running itself? this is not fantasy - it is practical reality in software development. Write a program to write programs; allow, then a degree of random error, outside the original algorithim, and see what happens. It doesn't work if you do it once: but try it a million times and you come up with bizzare new ways of doing things, often ways we simply don't understand. Chips are designed like this!

Obviously, this is evolution, with efficiency of the end product being the parameter of success. Whether it can generate intelligence is another matter entirely.

As for the robots on the island, they are an interdependent system: i.e one machine, not lots. There is no competition. What happens if the ore buckets start fighting to deliver their loads?

One might be tempted to say not much. Evolution has to start small: a complex system can't be the beginning, it can be the end. In the case of the great extinctions, it probably caused the end.
Magi

 

[kedaman]

There are indeterministic algorithms but they are algorithms nevertheless. We may not understand what a neural network or a genetic algorithm is working on but we understand how it works, it doesn't deviate from that. If we take the time to study it, we will see the algorithm has just done what it was suppose to do in a given situation. Randomness is not an algorithm, its input.

Evolution cannot produce complex systems; if there are such then they were created such that each component works for and does not deviate from the purpose of the whole as vice versa, the whole works for and does not deviate from the purpose of its components.

 

[TheMagi]

Evolution cannot produce complex systems; if there are such then they were created such that each component works for and does not deviate from the purpose of the whole as vice versa, the whole works for and does not deviate from the purpose of its components.

On what grounds do you say that? I am a young earth creationist, so I possibly don't disagree with your fundamental position.

A complex system can arise when one thing depends on another; it can be the case (obviously) that something can begin as opportunist, and end as dependence. I talk to my friends because I like them; yet I can come to depend on them, and they on me. If I do, and we together come to depend on something else, and so on, you get a complex system. That isn't design, it is emergence. It is a standard natural process visible in everything from evolution to the stock exchange.
Magi

 

[TheMagi]

Randomness is not an algorithm, its input.

Which, I suppose is the point. I have been trying (failing?) to say that evolution is only evolution when there is input, that algorithims become more than algorithms when they have somehting added. There is an algorithm which deals with the addition of input - that is what I was talking of.

Magi