The most insane idea I've ever had.

This is the most insane idea I've ever had, about anything. If you would like to post your thought, go ahead, that'll keep the thread alive. What I'm really wondering is if a programmer could actually do this or not...

The ultimate picture generator.
Other names:
The best screensaver in the world.
SETI@Home for pictures.

Terms:
800x600 resolution
16 colors
(Terms can change, I'm just trying to keep the numbers as low as possible so you can conceive my idea)

Make a program which, at the press of a key, will take every monitor pixel available (in our current terms, each pixel in the 800x600 pixel matrix) and assign it one of the 16 colors.

So, how many pictures can you make with this? 800x600^16
And on my computer calculator, that's a really big number.

And of course it's a big number! Think of how many pictures you can make on a 800x600 surface with only 16 colors. Then think about what you could make with even more colors, all the way up to 64-bit (they have that now, right?).

You could see what's inside of a black hole. You could see your future children at the age of 60. You could see SL 2

The possibilities are endlessly insane. If the same program could do this and figure the picture is recognizable (aka: something that humans could identify as... identifiable) then it would be like SETI@Home for pictures.

So, anyone up for making this program? Anyone still confused? Anyone calling up the mental institution and commiting me?

PS: If anyone really does embark on this adventure, I suggest...
Customizable resolution, customizable color depth, and an algorithm that can assign every picture (aka: pixel/color combo) a specific number/serialcode (no matter how farkin long it is) so people can see if you've found somthing incredible.

/insanity

So you want a screen saver that is just a completely random picture, and you want to be able to save the bit map ( .bmp ) of it?

The number of images possible on an 800 * 600 * 16 color monitor is 16 ^ (800 * 600). Which is friggin huge. (It is this way because htere is 16 options for the first pixel, 16 for the next... So if you only change 1 pixel you got 16 options, change 2 and you have 16 options for each color of the first pixel of which there is 16 options. Thus 16 ^ 2 ... etc. 800 * 600 ^ 16 would be a 16 pixel display with 800 * 600 possible colors).

It should be fairly easy to make but not very interesting. The thing with pictures is they do have some sense of pattern, some sense of continuity. Now to program that in there, to increase the likely hood that *any one* would *ever* get a recognizable picture would be harder of course.

Another option would be to do a full screen of completely random text. And if anyone gets one the recognise as a page of shakespear they submit it.

The set of all random and meaningless 800x600 images is larger than the set of almost meaningful images by such a large factor that were every single desktop to be running the screensaver as fast as they could (even with clever non-duplication) then it would take longer than the age of the universe to encounter a single meaningful image.

If you were to put a dozen monkeys in a room with unlimited supplies of crayons, pens, and paper then you would have better results within 30 minutes.

HAHAHAHAHA

you dont need monkeys...just come to my office.

btw dave-----------WHAT?!?

Why the hell are you calling the next version of Second Life SL2? Its definatly going to be Third Life!

Wrong James.

Its After Life.

wouldn't that look just like turning up the "Add Noise" filter in photoshop all the way? i mean if it's random and all..

Well, let's run the numbers.

We'll start small, just generating forum avatars. Let's generate all the monochrome (1-bit) 32x32 icons.

Sure, they'll be B&W, but because we are generating all possible icons, the very very best ones are garanteed to be there.

Let's assume that we've got a special icon generating CPU, who's only instruction is "Generate the n'th 32x32 1-bit icon, and decide if it's good or not" (I'm sure we can get one from the store). and let's say that this is a 5000000000 GHz chip, that is, it can generate 5 * 10^9 * 10 ^9 (5 billion-billion) icons a second.

So, if we start the machine going now, we only have to wait

5.7 * 10 ^ 282 Years for it to finish.

Then, when that's done, we can start working on Dave's 800x600x16 image generator.

First of all, sorry about the math... It was 4 a.m. what do you expect.

Secondly, I know the numbers are huge, and the chances of seeing a picture are pretty much impossible, but that's why I said it's the most insane idea I've ever had, okay?


But while we're on the subject of insane numbers, I remember some Linden saying there are more ways to customize your avatar with the available sliders than there are atoms on Earth. So a couple questions here.
How many ways are there to customize your avatar.
How many atoms are on Earth.

Big numbers scare me.

I know, I was kidding


The number of possible different avatars (discarding clothes and tatoos and suchlike) is the <inhale> product of the size of the range of all the sliders.

That is, if there were 4 sliders, and each slider had a range of 30 possible values, there would be
30 x 30 x 30 x 30
possible avatars.


Seven, but they're _very_ big. It's a secret particle physicists don't want us to know.

so since each slider has 100 options does that mean the total number of different avatars is 100x100x100x100... as many times as there are sliders? that's....big. you might as well just make up a name for that number. i'm going to say there are a million billion scrotillion banana-fana-fojillion possible avatars.

100 ^ X where X is the number of sliders in game. Or 10 ^ (2X) for more standard notation.

There is about 6.0221367 x 10 ^ 23 atoms in a mole right? A mole in weight is the atomic weight of the element grams, ya? Cuz a mole of C12 (carbon 12) is 12g.

Now thats a lot of atoms in the earth.

However there are 11 sliders just for the shape of the head. Which is 10^22 options. So its quite possible, especially once you take into account textures.

However the "real" number is a little smaller. Because are you going to be able to tell the difference between two AVs with all the same sliders except ones nose is 1 number bigger? Nah, those are really the same AV. So what range of difference is needed to differentiate between two AVs?

I was actually planning on doing this with text, Dave.

(To prove the infinite monkeys theory).

Have distributed computers start creating every possible combinations of letters, numbers, and special characters possible to infinite length.

When we're dealing with numbers like a million, five is pretty much six so we can just say those are the same number, so we really don't have a million numbers between one and a million.

Not even what I was talking about, I don't think. That or you confused me.

I am saying that I seriously don't think you could tell the difference between two AVs with only a single point difference. Is their hair a little longer? Their nose a little shorter?

Now you may not be able to tell one off from one slider, or maybe even 5, but I would bet that two AVs next to each other with every slider nudged by 1 or 2 .... they would still look similar but you could probably tell they were 'off' from each other. Just because of how much everything effects everything else.

Ah here is a good way of saying what I think I'm saying:

While there may be 10 ^ (2 * NumSliders) different AVs possible, the number of distinct AVs available is significantly less. Like maybe 10 ^ NumSliders, give or take.

Still a whole heck of a lot. But less.

The number of different avatars I could count and the number you could count are most likely different. It's subjective.

What I meant by my post is that "distinct" doesn't mean much -- three thousand seventy six marbles looks a lot like three thousand seventy seven marbles, but they're arguably distinct numbers.

Fair enough.

This thread reminds me of a conversation I had recently with a friend, insofar as we're almost talking holographic principles here....


I.E. the storing and retrieval of more information than would normally fit in a specific-sized flat area.

My friend and I were reading an article on a website about the theory that everything we see could be a holographic projection of a flat surface (membrane).

We looked at my groovy holographic mousemat (one of those plastic ones with lots of smileys at different depths that move when you look from side to side) and discussion ensued. We both agreed it could be feasable - at least we couldnt think of any reasons why not, when my friend popped the question: "So, what are black holes then?"

I thought for a moment, then said: "Theyre like inactive pixels that can't reflect light. Just like the spot on my mousemat where you placed your hot mug of coffee and RUINED THE BLOODY THING!"

too funny eddie, and that sorta stuff is very interesting to listen to!

I always wondered that if you could see everything that there is to see (resolution of wide eyes ^ all colors in the universe) and your brain actually processed it fast enough that you could live and see it all, if your brain would blow up or if you'd go mental or somthing randomly crazy like that.

I compare it to UCB's famous "bucket of truth". Whenever people looked in that thing they went nuts haha.

Minna (160,6 now holds a small version of a random image generator that I'm talking about.
And all it really makes is static But sometimes you can see images, if you... make it up real hard haha.

I also made a black and white version which is sorta neat but I think I'm gonna save it for a "What do you see" event.

And the idea morphs......

So here is a different but similar idea that combines this with another programming field that has always interested me - genetic algorithms.

So what you do is create a base set of 1,000 or so images. These are not quite random images, at least not on the pixel level. They are random shapes of random colors.

Now you create a website and get people to visit it. Participants play the role of determining fitness. They are presented with a set number of images (between 3 and 6 I think) and select the one to three they like best.

Based on this ranking you discard ones with too low a rating and 'mate' more often the ones with the higher rating. the mating process involves taking elements from each and merging them to produce two new images. And you need the chance of mutation.

The new images are subjected to the same ranking process. And images 'die' after 3 or so generations.

This is just a basic outline. It would need details like how exactly images are mated etc. Perhaps each image is divided into a random number of random proportioned pieces (between 3 and 9 or so) and swapped together to form two children. Part of the mutation may be the chance they are placed out of order. And there would probably need to be some smoothing done to the images after the merger.

That said it should produce some very interesting results.

Dave, are you certain it wasn't 4:20?

You dont need a distributed system to grep /dev/random =)
It's one of the things i do very often when im bored actually... See how long it will take /dev/random to generate my name, my girlfriends name, some cuss word... =)

Someone is already doing this with poetry. http://www.codeasart.com/poetry/darwin.html Website visitors are presented with two mutations and they select which they like best.

This paradigm is also being used to evolve electronic circuits. In fact someone has set up a distributed system like SETI is using where each computer has it's own local breeding population and it pereiodically sends its best designs back to the main system. Using this technique circuits have evolved that are superior to those pereviously designed by people for the same purpose, and natural selection has come up with circuit designs that people have already previously patented. Weirder still, some of the circuits that perform better than pervious human design aren't fully understood. They have extra parts that increase the efficiency of the circuit but no one can figure out exactly why.