One Thing That Real Evolution Can Do Which Simulated Evolution Can't

[jwu]

It isn't really that important to the E/C debate, but i found it pretty cool:

There is at least one thing that real life evolution can do which simulated evolution (genetic algorithms) or math in general cannot do: Forming a random number generator.


Generally this is quite a problem in math - there are pseudorandom number generators and so on which mimic real randomness, but as far as i know there is no real random number generator. Today in real life real random numbers are generated e.g. by measuring fluctuations in a resistor, by observing random physical events.

Well...some time ago there was a case that electric circuits which were made subject to evolution under real life conditions instead of a simulation evolved a radio receiver:
http://www.newscientist.com/article.ns?id=dn2732


In other words, real life evolution could create a random number generator by evolving a radio receiver, which uses a static signal (which is pretty much random) which it receives as a source of random numbers. In a simulation this couldn't happen. In real life, it is.

 

[ArnautDaniel]

You can (and people have) created random number generators along those lines (i.e. letting the computer use a a white noise type signal to generate a random number).

If you don't like that you could use a piece of radioactive material with its random signals to generate a random number.

It basically becomes a process of utilizing an external and natural signal.

It is just a matter of building the apparatus.

 

[jwu]

That's the point...it doesn't work in a purely mathematical environment but requires real life components.

 

[Chalnoth]

That's the point...it doesn't work in a purely mathematical environment but requires real life components.

Well, uh, so does a CPU, so what's your point?

 

[joey444]

Interesting.

 

[jwu]

Well, uh, so does a CPU, so what's your point?

The CPU emulates a purely mathematical environment, it is not part of the things which are simulated. It doesn't respond to outside stimuli.

My point is that in real life evolutionary systems can do something that they can't in simulations. As previously mentioned, this isn't a big deal - i just found it pretty cool when i found out about it (during a discussion on another messageboard where random number generators were relevant to the subject...so i may find this a bit more impressing than others ).

 

[ImmortalTechnique]

fascinating. i was actually wondering about this the other day. i used to be fairly adept at programming simple BASIC programs when I was a kid, and I was just wondering the other day how exactly the random numbers are generated. at first it seemed like it should be easy, but the more I thought, the more it seemed impossible.


good to know

 

[Maxwell511]

Well...some time ago there was a case that electric circuits which were made subject to evolution under real life conditions instead of a simulation evolved a radio receiver:
http://www.newscientist.com/article.ns?id=dn2732

It seems to me that the experiment was a bit flawed. I would have to see the paper on it. However I am reasonable sure that an amplifer congifurations is more probable then that of an oscillator since an amp only requires one transistor an oscillator requires at least two. They should have shielded the experiment from EM radiation.

In other words, real life evolution could create a random number generator by evolving a radio receiver, which uses a static signal (which is pretty much random) which it receives as a source of random numbers. In a simulation this couldn't happen. In real life, it is.

I don't think this is true, you could have simulated the line that was acting as an antenna with a simple oscillating current source and probably have gotten the simulation to produce the same results.

 

[jwu]

I don't think this is true, you could have simulated the line that was acting as an antenna with a simple oscillating current source and probably have gotten the simulation to produce the same results.

The simulation however requires a source for randomness of the simulated physical effects, causing an infinite regress.

 

[XTE]

There is nothing random about me.

In a .....RULES OF NATURE kind of way.

If you can't see it now, then you might never have hope.

 

[Maxwell511]

The simulation however requires a source for randomness of the simulated physical effects, causing an infinite regress.

The result is not random. If I was to hazard a guess I would say that it was a radio reciever tuned to a frequency of 60 Hz.

 

[jwu]

The result is not random. If I was to hazard a guess I would say that it was a radio reciever tuned to a frequency of 60 Hz.

In the example which i used as evidence that a radio receiver can evolve that was so, it got an external signal from that frequency. But in the same way a radio receiver can evolve which is tuned to a frequency on which there is just random static.

 

[Maxwell511]

In the example which i used as evidence that a radio receiver can evolve that was so, it got an external signal from that frequency. But in the same way a radio receiver can evolve which is tuned to a frequency on which there is just random static.

Random static doesn't exist on one frequency. It is a signal that has (or approximately has) all frequencies at the same magnitude, for example in the visible bandwidth the colour white is random static. The radio could evolve to allow any bandwidth and find random static.

 

[jwu]

Random static doesn't exist on one frequency. It is a signal that has (or approximately has) all frequencies at the same magnitude, for example in the visible bandwidth the colour white is random static.

...which also results in random readings on a specific wavelength if there is no overlying signal which outweighs this.

This is not really relevant to the point though.

 

[ArnautDaniel]

That's the point...it doesn't work in a purely mathematical environment but requires real life components.

No that's not your point.

In a purely mathematical environment random numbers are certainly possible.

The problem is constructing a physical computer that can generate random numbers, which is an engineering problem.

So all you are telling me is that a computer not equipped with hardware able to generate random numbers is not able to generate random numbers whereas biological systems with biological "hardware" able to generate random numbers are able to generate random numbers.

Actually it isn't much of a point you have there.

 

[Chalnoth]

Well, there are routines for generating pseudo-random numbers that are, for all intents and purposes, good enough. I don't think there's a real, significant difference between pseudo-random numbers and real random numbers, provided the algorithm is good.

 

[jwu]

No that's not your point.

Are you seriously trying to tell me what my own point is?

In a purely mathematical environment random numbers are certainly possible.

I'm curious...how would you generate random numbers in a purely mathematical environment?

The problem is constructing a physical computer that can generate random numbers, which is an engineering problem.

Actually that is rather easy - but it doesn't retreive those random numbers from math.

Actually it isn't much of a point you have there.

I even said in the opening post that it's not a groundbreaking issue.

Well, there are routines for generating pseudo-random numbers that are, for all intents and purposes, good enough. I don't think there's a real, significant difference between pseudo-random numbers and real random numbers, provided the algorithm is good.

Yep, but they only appear to be random...that distinction is what my point is about.

 

[Ondoher]

Well, there are routines for generating pseudo-random numbers that are, for all intents and purposes, good enough. I don't think there's a real, significant difference between pseudo-random numbers and real random numbers, provided the algorithm is good.

Pseudorandom number generators are predictable. This is actually the opposite of random. In terms of security applications, this difference is huge.

 

[Kripost]

The root of the problem is that real computers are deterministic, and have finite memory which implies there are only a finite number of states.

 

[jwu]

Same question then - how would you generate a random number purely from math or algorithms but without real life influence?