Why Auto-Evo is Dead

First off, I know that this is going to piss off alot of people. But I don't care, because this has been bugging me since day one.




1. Why the game must stand up without Auto-Evo

Okay, lets begin by doing what everyone seems to be doing at the moment, and quote a popular comedic game reviewer, Ben Croshaw, netter known as Mr Zero Punctuation.

"A game must stand up on single-player alone."

Why? Because there will always be factors in the way of the "multiplayer experience"

Such as lack of internet connection. Your local exchange can and WILL go down.
Or young people screaming into the microphone every 5 seconds that you have to mute EVERY SINGLE GAME (I'm looking at you, call of duty.)

And what is the factor that prevents us from relying on Auto-Evo?

We Don't Have an Auto-Evo system.

You can't build a game around a system you

a) don't have
and
b) have no idea how to implement.

It means that we'll be sitting on a game that is broken forever because we haven't inserted the fnal piece of the puzzle, the piece that will never exist.


Which brings me to two.




2. Why wew can't create Auto-Evo

Now, i tried explaining this giving people credit, but it glazed over alot of peoples heads for some reason so i'm going to explain this again, very slowly.


Algebra 101.

Lets say we have the following:

2 + x = 3

Solve for x.


Now, many of you will be going "Oh come on i'm not stupid x is 1."

and you would be correct.

now take this:


2 + y / f + g * j + e / 7 * x * H ^ L - ((4-s)+q*l) * z / ( 80 + w) = 3

solve for x

Much harder?

Of course it is. in fact, it's pretty much unsolveable, because there's likely multiple possible answers.

How does this relate to auto-evo?

Well, Every piece of software in a computer, video games included, are just math.

The CPU is a very very fast calculator, and nothing more.

An auto-evo algorithm will have to be something like the above, however,

we're not just solving for x.

We're actually producing something like that.

2 + y / f + g * j + e / 7 * x * H ^ L - ((4-s)+q*l) * z / ( 80 + w) = creature.

but we have to find out just exaclty what equals creature. that means not only placing a,b,c,d,e,f,g,h,i,j in an equation, but placing them in the right places and performing the right operations on them.

Let's go back to the basic equation

2 + x = 3


lets say we get x wrong.

x=40

2 + 40 = 3?

no, no it doesnt.


What if we get x right, but have the wrong operation?

x = 1

2 - x = 3?

no.

If we get anything wrong in developing an algorithm for auto-evo, the answer will be wrong.

And not just a little wrong. Due to the fractular nature of such an algorithm, a small error at the left side of the equals makes a huge difference at the right side of it.

I previously challenged people to come up with a auto-evo algorithm,

and many, many times all i got back was a mere description on how it should work.

This. Does. Not. Help. Me.

I can't just say to the computer "Yo computermanbox thing. I want an evolution system and it should have stuff like the legs changing according to how they run and stuff. think you could cook some-a that sweet sweet code pie i like?"

It doesn't work that way.

And i can see alot of people getting insult and saying "I KNOW all this, Bashi. Stop treating us like idiots."

But if you insist on arguing with me on the subject of Auto-Evo, then you obviously don't. I cannot wave a magic wand and bend maths to my every whim. I'm a good programmer, but that is all that I am. I don't have degrees in quantum bloody physics. I talk in the computer's language. That's all. Anything I don't know yet, i learn on the fly. But learning something such as an Evolution algorithm working at the level that is being requested? How am I meant to learn about something that doesn't exist yet?

If i were able to create such a system, that works the way described, I would be a billionaire.

Not only that, we have a couple more problems.

The first thing is, as you stated, that the whole auto-evo would be a biiig mathematical function. It'd have several optimal spots, to which creatures would converge eventually. This would seriously hamper diversity as you'd see zounds of similar creatures because their stats are the best. Of course enviroment would help by saying which of those optmial solutions would be better, but still. If you want a real life biodiversity you'd need to input so many variables and so many systems (take a quick read on how scientists now think viruses are shaping evolution by moving genes to and fro cells) even a modern supercomputer would probably be unable to simulate.

Second, that computers are dumb. Plain dumb. They have no aesthetics sense. They'll make monstrosities with the weirdest leg angles, body size and organ dispositions because they have no sense of what looks good or not.

Djohaal (Thu 16 Dec 2010 - 4:27) wrote:

Second, that computers are dumb. Plain dumb. They have no aesthetics sense. They'll make monstrosities with the weirdest leg angles, body size and organ dispositions because they have no sense of what looks good or not.

Neither has nature.

----- ----- ----- ----- -----

This is exactly why only partial evolution, if at all, will happen. Editors will always be a part of the gameplay.

Bashinerox wrote:

text...

What angered you?

Commander Keen wrote:

Djohaal (Thu 16 Dec 2010 - 4:27) wrote:

Second, that computers are dumb. Plain dumb. They have no aesthetics sense. They'll make monstrosities with the weirdest leg angles, body size and organ dispositions because they have no sense of what looks good or not.

Neither has nature.


This is exactly why only partial evolution, if at all, will happen. Editors will always be a part of the gameplay.

There's an anthropic component to why we find things nature evolved here pretty. We evolved to think so

Alright, given we dont have auto-evo, what are our possible options? What is easiest, and second easiest thing to code? Do you have enough motivation to do it?

Bashinerox wrote:

I can't just say to the computer "Yo computermanbox thing. I want an evolution system and it should have stuff like the legs changing according to how they run and stuff. think you could cook some-a that sweet sweet code pie i like?"

Change legs according to how they run? I thought auto-evo was supposed to be random, not what is basically the AI going into the editor and changing it for you...

What is the current auto-evo concept? Darwinian, or... This other one?

The Uteen wrote:

What is the current auto-evo concept? Darwinian, or... This other one?

It's called Lackardian, methinks. Our current auto-evo concept revolves around a hybrid system.

Darkgamma wrote:

The Uteen wrote:

What is the current auto-evo concept? Darwinian, or... This other one?

It's called Lackardian, methinks. Our current auto-evo concept revolves around a hybrid system.

Isn't it Lamarkian? Anyway, I thought that was direct editing, so I didn't want to say that...

But why not Darwinian? A hybrid definitely will be more complicated.

True

x=e/(1-y/f-g*j+((4-s)+q*l)*z/(80+w))/(h^l*7)

So, I knew that Auto-Evo was going to be very hard to make and very simple in it's final form. Simple to the point of have an only tangencial relationship with actual evolution.

But none?

As in, nothing?

This is what we have designed this whole project around, the idea that we could create a system that allowed organisms to change and "evolve" according to their environment. Now, I'll say this: I'll give up plants. I'll give up being able to actually have an editor. I'll give up sentince. All if you can just get the most basic, tiniest random change to occur.

Bashi, I know this is hard. I know this is horrible, horrible work. I know how much heart and love you're putting into this. I know most of all how stupid it's been with "Auto-Evo will solve it." But this is the most basic, fundimental core of Thrive. This is what brought me here, and I'm sure it's what brought many others here as well. This is what makes the game what it is.

You can forget about player actions if you need to. You can forget about everything. Just please, please tell me that we can make something. If you're not satisfied with what we tell you, tell us. When you need an algorythm, ask. But please, please, please don't abandon this part of Thrive. Because this is Thrive. This is all we need. If this is gone, than everything we've worked for over the years will have been nothing more than a mental excersize.

And lastly, I'm sorry. I'm sorry I never caught on to your struggling earlier. I'm sorry for all the times I've been lazy or incompitent or slow. Please, forgive me. Forgive us. We need this. We need you. And I'll venture to make this whole thing a joke to be put in the next dev blog and say this: I love you.

Tenebrarum wrote:

-longpost is long-

No need to get that emotional tene. I think it is possible for we to find a satsfactory middle-ground for evolution in thrive, but I don't think it'll be anything similar to automated darwinian evolution...

Djohaal wrote:

Tenebrarum wrote:

-longpost is long-

No need to get that emotional tene. I think it is possible for we to find a satsfactory middle-ground for evolution in thrive, but I don't think it'll be anything similar to automated darwinian evolution...

I know that, but what I hear when I read this is that evolution, as in, organisms changing to suit their environment without need of human input will not be a part of this game.

That's what I cannot take.

Also, just so everyone knows, I've been in contact with a man who's been working on a darwinian evolution simulator for an extended period of time now. When a first asked him, about a year ago, he said he wanted to join but lacked the time. I've just asked him for help and I'm waiting on a response.

To be honest, I never really saw how Auto-evo would pan out. I mean, if there was a way to accurately simulate evolution, it would have been created years ago, and for scientific purposes, not for a game. We're going to have to think of alternatives. Obviously, it cannot be ridiculously realistic for the reasons that Bashi pointed out. However, it can feign realism. There are few players out there who will look at this game and say 'Pish Posh! This isn't realistic!'. Of course, I have no coding background, so I may as well shut up, since I have no clue how to implement it. Also, I'd recommend coming to the chat so we can discuss in real time. Djohaal and I are already there.

fireballs619 wrote:

To be honest, I never really saw how Auto-evo would pan out. I mean, if there was a way to accurately simulate evolution, it would have been created years ago, and for scientific purposes, not for a game.

But there ARE darwinian simulators. The problem with is is that it takes extremly long for any significant change to occur. Often, running them for the whole night at 32x time compression makes nearly invisible changes.

Also, sorry, no time to join the channel right now.

fireballs619 wrote:

To be honest, I never really saw how Auto-evo would pan out. I mean, if there was a way to accurately simulate evolution, it would have been created years ago, and for scientific purposes, not for a game. We're going to have to think of alternatives. Obviously, it cannot be ridiculously realistic for the reasons that Bashi pointed out. However, it can feign realism. There are few players out there who will look at this game and say 'Pish Posh! This isn't realistic!'. Of course, I have no coding background, so I may as well shut up, since I have no clue how to implement it. Also, I'd recommend coming to the chat so we can discuss in real time. Djohaal and I are already there.

QFT. It doesn't need to be real, it must seem to be real

Djohaal wrote:

It doesn't need to be real, it must seem to be real

Yes.

Now tell me how.

You see, the way it works in real life is arguably the simplest way out there. We can't do that. We can't do Lamarkian, where we'd have to program each "thing" the organism can do and the sort of effects that would have on it's structure. What can we do?

Have the game open the editor, do something random, save and close?

Repeat for each organism that's born?

But that might cause problems if every organism is unique...

Another big problem is speeding it up.

Tenebrarum wrote:

Djohaal wrote:

It doesn't need to be real, it must seem to be real

Yes.

Now tell me how.

You see, the way it works in real life is arguably the simplest way out there. We can't do that. We can't do Lamarkian, where we'd have to program each "thing" the organism can do and the sort of effects that would have on it's structure. What can we do?

I think programming it to count every "thing" an organism can do would be an order of magnitude simpler than auto-evo still. Also chat is alive now.

Tenebrarum wrote:

If you're not satisfied with what we tell you, tell us.

He just did, no?

Anyway, I have absolutely no coding experience, and I am an absolute idiot when it comes to math (Exam's tomorrow. I'm fucked.) so I can't help. Of course, even if I were Albert Einstein, I would be completely unable to write this equation.

Which is why we will simplify- create a new system which will not cause Bashi's brain to implode.

Hmmm.... I'm blank, as usual...

To be honest, I'm stuck.

All progress on the engine has halted yet again because the decisions made on this whole auto-evo thing decides everything.

It determines the heirachy of information in the creature engine
It determines what if any physics engine we choose (there are many types of physics solvers)
It pretty much outlies the entire framework of everything.

And that means that I cant just code for something that doesn't exist, because that's like.. building the interior of a car that hasn't been designed. how big do the chairs need to be? how do they attach to the bottom of a car? what shape does the dashboard need to be?

and so on.

The framework for getting something playable up soon is half-built. But half-built gets us nowhere, I cant run half a piece of code.

Which gets us to a point, we need to find consensus on how to implement evolution* somehow.

*As in the game mechanic, not the real mechanics as bashi already started it is too insane.

Djohaal wrote:

Which gets us to a point, we need to find consensus on how to implement evolution* somehow.

*As in the game mechanic, not the real mechanics as bashi already started it is too insane.

In great detail.
Everything needs to be mapped out, not just

"well it should do this and this"

And i don't mean respond to that with

"well it should do this and this and this and this and this"


HOW does it do "this and this and this and this and this"?

Bashinerox wrote:

HOW does it do "this and this and this and this and this"?

I've already explained why I can't do that, so I might as well shut my mouth.

It appears that we are momentarily, dare I say it, fucked.

"Efforts and courage are not enough."

InvaderZim wrote:

Bashinerox wrote:

HOW does it do "this and this and this and this and this"?

I've already explained why I can't do that, so I might as well shut my mouth.

It appears that we are momentarily, dare I say it, fucked.

"Efforts and courage are not enough."

We aren't fucked, ..wait, the censor's been turned off.. Anyway, we aren't fucked if we decide to turn towards editor-centric gameplay and drop auto-evo for now.