Friday 12 February 2016

make sense?

Victor Nava You know a few months ago, I was working on a program to solve permutations and combinations.

This is a very interesting problem because you can think of many things in terms of combinations or sequences. Put this right thing in the right place and it makes something. Somehow, Music is a sequence of notes, books are sequences of letters, a painting is a sequence of material.

This somehow got me thinking about infinity. What is infinity, is the universe infinite? is anything infinite? is infinity possible?

Then I stopped thinking about that.

A few months later I was working on different program that generated drawings. A sequence of pixels on the screen.

Than I got thinking about combinations and infinity again. You know when you are not working you have time to think about silly things.

I asked my self, how is it possible that a program can make a drawing? well that's easy, it is possible because I am telling the program what to do. Cool. Then I thought what if the program could make drawings with random parameters until it makes one that I like? umm that sounds possible.

Than I though, what don't I just make a really dumb program that generates ALL of the possible drawings I can fit on my screen. Shit that would take a long time, but just for fun, how many drawings can in theory be drawn on my computer screen?

My initial answer was: a infinite number.

That can't be right, if the number of pixels on my screen is finite and the number of colors that each pixel can be in at any given time is finite, how is it possible to draw an infinite number of pictures?

Let's see, my screen has a resolution of 1200 by 900 pixels. That means that there are in total about 1 Million (1200 * 900) pixels. Each pixel can have lit in 16777216 different colors.

So how many drawing can be drawn on my screen?

The answer is simple and it comes from combinations.

An image is nothing more that a sequence of numbers, that represent pixels.
Each pixel can be of a particular color. That's it get the right pixel lit up with the right color and you get an picture.

To calculate the number of the possible pictures that can be shown on a screen you take the number of pixels and the number of colors that each pixel can be in and apply the mathematical formula to find the number of combinations of two sets.

The result is a very very very big number. However it is a finite number, which means you can only put so many different pictures on a particular screen.

Hold on, there are so many things I can draw on my screen.

Realising this made me a bit sad.

I started thinking about creativity and imagination and soon realised that this terms as romantic as they sound, obey the same rules. There are only so many things we can think of. Because the number of cells in our brains is finite and thoughts come from our brain, from the combination of signals from neurons firing messages to each other.

So in theory, you could simulate every single cell in a brain, and run a program that calculates every possible combination of connections, and messages. The same as with a computer screen.

This sound impossible, and it is at the moment.

To generate all of the possible pictures on my screen, it would take my computer an enormous amount time. But if you give it enough time it will eventually get there. Now, this is using current technology and dumb algorithms.

The other interesting question is, how many pictures that make sense can be shown on my screen?

Again the number is finite. Because this is a subset of all the possible pictures. So in theory it should be easier to find the answer, right?

No. Doing this is actually really hard. Because to do this you need to tell the computer what a picture that makes sense looks like. And we don't know how to describe that in words. We can tell what makes sense but we can't figure out the formula.

However, you can teach a computer how to find things that make sense by training it. Pretty much the same way you teach humans, by example. You show it a bunch pictures that make sense and a bunch that don't and let it know which one is wish. If you give it enough examples it can "learn" to recognise the pattern. This is already possible, the problem is how to find the right data.

But in theory you can take a computer programm and feed every single web page on the internet, every sms, facebook and whatsapp message, plus all the books ever written. And calculate the right combination of letters to form words, sentences and stories that make sense. Within the number of stories that make sense there will be many that will be superior to any other ever written.

Pause here for a moment. This programs are not taught how to write, they are taught how to recognise patterns that humans like and emulate them. To the computers this patterns are just numbers. Not pixels, or letters just numbers.

There are a finite number of letter in the alphabet and a finite number of letters you can put in a 400 pages of paper.

A book is a combination of sentences.
A sentence is a combination of words.
A word is a combination of letters.
All finite. And predictable.

Where the ideas to combine letters in an way that makes sense comes from is a mystery.

But when you can see all possibilities you don't need to imagine. You just need to select.

From all of the possible books that can be written lie the works of Dostoevsky. Computers will eventually generate all of them, plus the ones he didn't have time to write.

They will be able to imitate the style of writers. And combine styles from different people to form new styles. And we won't be able to tell the difference. Or maybe we will, we'll say things like: "this book is too good, it must have been generated by some stupid program".

to be continued...





Me:
Where the ideas to combine letters in an way that makes sense comes from is a mystery.

sense: A sense is a physiological capacity of organisms that provides data for perception. The senses and their operation, classification, and theory are overlapping topics studied by a variety of fields, most notably neuroscience, cognitive psychology (or cognitive science), and philosophy of perception

So: If we can map out what make sense to us and we can formalize this somehow so the computer can understand they we can solve this mystery

But how we can do this?

Sense: to me its just a sequence of the mind electricity transitions that its happening in our mind.
But what formula we can use here to save it when "something makes sense" to us so the computers can understand?

Another thought:
Is this the right approache? Or do we have to save to computer what are the facts about the elements around us and what is this that we want so a situation wil make sense. What i mean is:

Is it the "make sense" a process based on existing defined truths and the results that we want?

results = :foo

defined_truths = [4,2,3,5]

is the process make sense

defined_truths.include? results

the actually "include?" method when its happening do we do the same with our minds?

So: is it the "make sense" a sequence of examples that they can work? and it can aply to all the stuff in life?

results = i wanna fly
defined_truths = [gravity, weight]

the make sense

If i wanna fly the gravity should be less that x
Can i do this? yes ok makes sense

So for me its an array of examples that can be true to nature laws and the laws we define smile emoticon

and again it comes something else:

Who can define laws and why and how? Based again in a array of valid and accepted examples?

Big story and very interesting grin emoticon




i was walking now and i was thinking that we have to DEFINE everything to the computer so its common sense to them also.  What it is each Object and what aretheir attributes so we can compare select etc




No comments:

Post a Comment