• Bleach: The Thousand-Year Blood War - Part 3: The Conflict is expected to arrive in 2024. Until then, check out the latest news from the Bleach Anime Section: HERE! ~ The BA staff.

Outlier's Colourings and Graphics

xx

  • xx

    Votes: 1 50.0%
  • xx

    Votes: 1 50.0%
  • xx

    Votes: 1 50.0%

  • Total voters
    2
  • Poll closed .

NeoPlatonist

Life is a song cue . . .
Joined
May 12, 2013
Messages
902
Reaction score
1,296
Points
4,041
Duuuuuuuuuuuude,

I'm so happy you found a place where you can host your videos without fear of them being blocked for Copyright. It's not like you want to monetize them and I'm grateful to get to view them. :thumbs
I'm in agreement with Polgarena - that Bleeeeeeeeeeeach-Stretch Siggy is veddy veddy coolio. Me likes it a lot.

But those videos! How do you ever conceive and execute them. Just. Wow. The black and white manga opener - so pixelated in a moody atmospheric way that really sets a tone. And your edits and quick changes into new scenes, overlapping dialogue and music are just wonderful. What a creative fellow you are.

But of course - my fave - was seeing that slowed down battle with special added color highlights. What a treat. Sometimes I (and I know a lot of other people do this too) will go through a scene frame by frame just to savor the moments. It's so fun to see it re-imagined and underlined with vibrant colour. All of this was a real treat, and I truly enjoy when you share your talents with us. Thank you.

Now.... if only Polgarena had some time for more drawing...... :jk
 
Joined
Jul 22, 2014
Messages
1,281
Reaction score
1,241
Points
2,641
Duuuuuuuuuuuude,

I'm so happy you found a place where you can host your videos without fear of them being blocked for Copyright. It's not like you want to monetize them and I'm grateful to get to view them. :thumbs
I'm in agreement with Polgarena - that Bleeeeeeeeeeeach-Stretch Siggy is veddy veddy coolio. Me likes it a lot.

But those videos! How do you ever conceive and execute them. Just. Wow. The black and white manga opener - so pixelated in a moody atmospheric way that really sets a tone. And your edits and quick changes into new scenes, overlapping dialogue and music are just wonderful. What a creative fellow you are.

But of course - my fave - was seeing that slowed down battle with special added color highlights. What a treat. Sometimes I (and I know a lot of other people do this too) will go through a scene frame by frame just to savor the moments. It's so fun to see it re-imagined and underlined with vibrant colour. All of this was a real treat, and I truly enjoy when you share your talents with us. Thank you.

Now.... if only Polgarena had some time for more drawing...... :jk

Well, the videos I did with previous opening songs with different colour themes were all deleted by youtube because of the song. Veed.io was a temporary solution and it does not let upload videos in higher resolution. Now, I have made my own personal website where I can upload whatever I want, but it's for another purpose and posting there would not have made sense. Another solution is to just embed it from Google Drive or OneDrive.

About the "BLEACH Ichigo" signature, "BLEACH" never existed. I painted the reaitsu again to be like "BLEACH".

About the videos and their editing, I don't use a fancy editor. Any basic editor would work for me provided there are colour management options. I mainly play with colours and that's my favourite fun thing to do. People use Adobe products very enthusiastically and their editor is considered one of the best if not the best, but I have never used it. I do change editors from time to time in quest for simplicity. My current one is:
It's free and fun to use.

About drawing, there is another fanart I am working on. Funny thing is I wanted to do it for Zangetsu but ended up establishing Yhwach. It's far from complete, but I have established the basic lines and overall basic concept of him. To give the teaser, here is my work in progress:
OO2GpOH.png

I would say it's 5% complete yet. Even though the shared image is a raster, this would be my first vector work here. Vector means if I wanted to make a billboard out of it, it can be printed without any worry because it's not a pixel work.

I am using "Infinite Design" on my phone which is a simple vector application. Even though there are vector brushes in other painting applications (like Concepts or ClipStudio Paint), using them with pixel works create a hybrid work which I am not a fan of it. What's the point of it if it's still saved by pixel in the end? But there are lots of use cases for that kind of hybrid work. For my taste, I would either go 100% pixel or 100% vector. Concepts is a good app with great vector brushes, but if saved in svg which is my desired file, it removes the extra information from the brushes and that gives a bad shock in the end if someone wants a 100% vector work like me. Inkscape, Adobe Illustrator and Affinity Designer are also good desktop vector applications, but I wanted to do that with my finger on my phone and for that, Infinite Design is best for me. It's simple and clutter-free.
 
Last edited:
Joined
Jul 22, 2014
Messages
1,281
Reaction score
1,241
Points
2,641
I kept a promise to myself that whenever Yhwach vs Ichibei battle would be animated and in-colour, I would redo it in black and white, but with my own taste in it. I initially had a wish of the anime officially making the battle black and white, but it's alright. I shall bring my own wish to life by myself (just like Yhwach's voice regain lol).

Some fans have reservations about this battle. I had my own as well. I had read many comments and I tried to trim down the commonalities that what reservations they had. It boiled down to these:

1) Battle wasn't continuous and was switching to other Royal Guards battles.
2) Lack of music in later half.

While I was doing my personal wish of redoing the battle in black and white, I accommodated those two points as well.

With all those points, I started working on this project. I tried several possibilities until I was satisfied with the results. Music OSTs were difficult to put in the right timing. I had to listen and repeat dozens of time to match and fit with the timing of the visuals to see where they would actually fit.

About the visual result, last time I tried that was with the trailer and I tried to make that manga-like. But this time, I had a different visual theme in my mind. I also tried focusing on single colour besides black and white just like Sin City style, but I wasn't personally satisfied because it was giving unnecessary distractions. I discarded that. I also didn't disturb the flow of battle. For example, I could have added generated frames to slow-mo some scenes or could have added Toei-like redundancy, but I avoided that.

The following is a 15 minute uninterrupted complete battle with my personalized touch along with new OSTs. It's not that I simply converted it into black and white. Nah, it's not that simple because simply converting it onto black and white would never give this style what I personalized. I had to do lots of different things in different layers in quest for my personalized vision. Here it's in my drive (headphones are recommended):


I hope it's enjoyable for you as well it was for myself. The best parts of the battles are yet to come in cour 3. I have created my notes as well while doing this project so that whenever the remaining batlle would come, I would rejoin it with same visual experiment or perhapsdo something different depending on my mood. I hope we all survive till then in this world at wars...
 
Last edited:

NeoPlatonist

Life is a song cue . . .
Joined
May 12, 2013
Messages
902
Reaction score
1,296
Points
4,041
I just watched it - wonderful to see it all edited together. Thank you!

What a treat! :thumbs
 
Joined
Jul 22, 2014
Messages
1,281
Reaction score
1,241
Points
2,641
I have been getting lots of questions in my personal life regarding AI in art. I actually feel this topic to be dull, but because it's being asked so much, I thought I would make a post about it.

My background coming from Computer Science where I spent 1.5 years learning AI, ML and creating projects in them. In those 5 years of Computer Science education, AI didn't interest me compared to other Computer Science topics. For example, I loved topics like Automata Theory, Chaos Theory, Visual Computing, Logic Building and my most favourite of all, Data and Storage. But my final project was Chaos Theory in creating accident detection system. Although this topic wasn't my personal favourite, I only did that because the project copyrights would be owned and published by the institute and not me. That's the sole reason I picked that project topic and not my personal favourite. My work VCDS (Chaos Detection System) is published in the project book published by the institute and can be viewed in the institute's library archives along with physical implementation of the project in the disk because it's in the public domain. I would not explain 'V' because it would become too personal to share here.

I had to give this background to highlight some credibility in speaking on this topic. Because the word AI has become a household name, and not everyone knows the context of it, I felt the need to share some of my background.

Then about art, I have recently created and about to publish two of my finished games, all from scratch (non AI work), with the third one being a work in progress. All the artworks, the creative assets, the programming, the logic and gameplay elements were solely done by me i.e. from conception to finish projects.

If someone would ask me this question of AI in art about 12 years ago, I would have said that I might not be the suitable person to ask this question. Even today, I would say the same that I would not be the suitable person to ask, but my best friends persuaded me enough to speak on this.

To clear some things first, I would not include things that I have no knowledge of. For example, I won't discuss the copyright authorship issues as they are the domain of lawyers or relevant people. What I would be discussing here are the foundational aspects of AI in art from my mental perspective. My main objective is to bridge the hatred that has emerged between tech world and art world. Another thing I would not discuss is the economic model or monetary aspect associated with both domains. In the end, another goal of this post is to motivate people from both worlds, and not to spark a feeling of doom.

Let's begin...

When people imagine what AI is like, most of them imagine it some kind of a monster that's about to create havoc in their worlds. In other words, it's the fear. Take an example from BLEACH. When Gremmy fantasized the monster in his mind, that monster destroyed him. It was not actually the monster itself that physically destroyed him. It was his assumption of the power that scared him. His body couldn't hold the power.

When we talk about AI, we are trying to process it like how Gremmy did. Even when Kenpachi was not in the mood to kill Gremmy and just wanted to enjoy the fight, it was Gremmy himself that imagined Kenpachi to be a monster that killed him. This approach is similar to how we are viewing AI at this moment. Imagining it to be a kind of monster. It wasn't the monster that destroyed. It was the person himself that tried to measure the depth of monster and eventually got killed.

Many people since the dawn of time have attributed imagination to be the most powerful human faculty of the mind. Some even attributed that it didn't even come from mind but something other than that. Some attributed that it doesn't need to be perceived through senses at all.

Another important human faculty of the mind is the intelligence. Some attributed that it's the most powerful faculty human mind has. Does it mean it's important than imagination? This is the war between two fundamental faculty of human mind. Imagination vs Intelligence.

If we associate attributes to both, we often associate words like learning, reasoning, problem-solving with the word intelligence while we associate words like beauty, fantasy, dreams, creativity with word imagination. Often, there are words that are associated with both of these depending on the context. For example, the word idea.

If you would ask what's the definition of intelligence and imagination to 10 different people, it's likely that you would hear 10 different answers. That's perfectly fine because interpretations are subjective. Similarly, if you also ask what's the definition of AI and Art to 10 different people, it's likely that you would hear 10 different answers. That's also fine because interpretations are subjective depending on contexts.

From the origin of time to this present moment, it's flowing into the future. What you do or don't do, it constitutes an experience that is inherently tied to the perception of time. Experiences can incorporate various events. You can recall those as memories if you are able to. I would discuss it later.

In my mind, the three pillars on which Computer Science is standing are data, algorithm, and computation.
The word 'information' is closely associated with 'data' while the word 'calculation' is closely associated with 'computation'.

Raw Facts and Figures are called Data.

The word raw means that the facts have not been processed yet to get their exact meaning. The process of calculating data is called Data Processing. The result of data processing is "Information".

Let's dissect the words. What are raw facts and figures? Think of it as a sugarcane before being inserted into a juicing machine. A person who has not seen a sugarcane stick in their life could mistake the sugarcane stick as a bamboo. Raw is something that isn't manipulated yet. It's a representational value of subjective truth in the eyes of beholder that has not been processed yet to understand the relative truth that may come out as a result. A person who mistakenly believes a sugarcane stick to be a bamboo, there is a subjective truth in it based on the similarity of the shape. But when it's chewed or turned into a juice to drink, then the relative truth might come to lay claim with the notion that it was indeed a sugarcane stick and not bamboo. Regardless of human understanding or mistaken belief, the stick was always of the sugarcane. In other words, raw facts didn't cater to human reasoning.

Data may be jumbled, messy, not understandable, indistinguishable, but its representational value is irrespective of human meaning to it. It is 'as-it-is'.

The "P" in CPU (brain of computer called Central Processing Unit) stands for "processing," specifically, data processing. Processing data into information is the fundamental purpose of a computer.

Some data is not relevant or informational. This irrelevant data is called noise. For example, if you create an audio recording of a piano concert, you might hear people in the audience coughing, or the sound of a ceiling fan. These noises are irrelevant to the purpose of the audio recording, which is to record the sound of the piano only. The relevant data is the sound of the piano. It answers the question, "what did the piano sound like?" The remaining data (the noise) does not answer that question, so it can be ignored or removed.

The point to be noted here is that "Noise" is also "Data" even though we don't want it or it's deemed as irrelevant. But "Noise" can be relevant if our purpose is to record the behaviour of audience during a piano concert. Then, our playing of piano becomes the "Noise" and the audience audio becomes our relevant "Data". It totally depends on our needs and not to get fooled by the very idea of "Noise".

The separation of "Noise" from a signal is called signal processing. Similarly, data processing identifies meaningful data, and separates it from the meaningless data. The meaningful data is then interpreted, combined, modified, connected, and structured into something new called "Information". Information informs you of something. It answers a specific question. It represents a specific truth or fact. Data is the collection of recorded values from which information can be ascertained. For example, consider the question, "what is the temperature outside?" Data provides the basis for an answer to that question. If the data is "27.6" and "Celsius," the answer is, "Outside, the temperature is 27.6 degrees Celsius." You must know what "temperature" is, and what "degrees Celsius" are, to process the data into information.

Raw facts before processing may or may not be specific, but facts after "Information" are specific.

The following is just a simplified example of raw data, and how that data can be assembled into information:

Data:
UT, 1234, Shibata, West Rukongai, Soul Society, 8036663311, 84084, Yuichi

Information:
Yuichi Shibata, 1234 West Rukongai, Soul Society, UT 84084 (803) 666-3311

In this above instance, the original data was interpreted, organized, and formatted according to predefined parameters. Now the meaning of the data is clear: it is the contact information for a person named Yuichi Shibata. The words "format" and "information" are closely related. "Formatted data" is data "in a form." Its values are arranged to conform to a predefined structure or shape. It is "in formation” (i.e., information).

Let's come to algorithm. It can be called as "the way"....Consider a simple example. Suppose you want to go to airport, and you know the way to go there, there are specific steps that you would consider regarding the path. You might turn left, right, do a U-turn. The main thing is that you have a goal in mind and that is to reach there. You can follow a shortest path or a longest path or the convenient path. That all depends on your own context.

Algorithm is the area that incorporate logic, problem-solving mechanisms, a way to reach a goal or at least approach towards it. This is the area that involves programming which is based on pseudocode.

The third pillar is the computation. In essence, all computations involve calculations, but not all calculations necessarily involve the broader aspects of computation. Think of it as a pacman that wants to eat all dots. Before coming to that, let's go back some 50,000 years ago with earlier human having a sense of more vs less. As time passed, the idea of ancient more vs less later evolved to quantification. For example, in order to count animals, it is often guessed that the earliest method would be tally sticks or tally pebbles or tally marks. The method uses one to one correspondence. Items that are being counted are uniquely linked with some counting mark.

There is a modern example of that. In street cricket, suppose there are 6 people that are gathered together to play. But how are we going to distribute the turns in a way that is not quarrelsome. A seventh person who is not supposed to play the game is unanimously selected to do the tally. That person would mark 6 distinct lines making a one-to-one correspondence with the representational value of the link. For example, let's assume the triple dashes to be the line. Then,

--- 5
--- 1
--- 3
--- 4
--- 6
--- 2

The digits are hidden by the bat. The assigning of the digits totally depends on the heart and whim of the 7th person who is not supposed to play at all. The players are called to run to select their lines. They don't know what line would have what number. They claim the line by running towards the marks and putting a finger on the line to select. When all the players have placed fingers and claimed the lines, the bat is lifted from the hidden numbers and those values represent the turns they are supposed to go and bat. This is one of the most rudimentary examples I can give of computation that's easier to grasp.

About pacman (not to be confused with Package Manager of Manjero Linux Distribution), think of Pac-Man as the algorithm, and each dot it eats as a step in the computation process. The systematic way Pac-Man navigates the maze and consumes dots mirrors the step-by-step nature of algorithms executing tasks or solving problems in a computational context. In this way, we can think of computation as a resource eating facility. It's the same ancient idea of more vs less.

The more computation is required, the more resources are consumed. The relationship between computation and resource usage can vary depending on factors such as the complexity of the algorithms, the size of the data being processed, and the efficiency of the implementation.

We have now understood the three pillars in a very rudimentary way. Let's go and knock the door of AI. But before doing that, it would be a good idea to look at Turing Machine first.

Think of a Turing machine as a simple, imaginary machine. It has an infinite tape with cells. Imagine it like a reel of old photographs with squares where each square is a cell. The tape can be a symbolic representation of memory. A read/write head that can move left or right (one cell at a time). Imagine this to be a lamp that can throw spotlight on a single cell only and has the ability to detect/read what's inside the cell and also has the ability to erase and rewrite new thing inside the cell. Imagine some rules. These rules tell the machine what to do based on the symbol it sees on the tape and its current state. It reads a symbol, writes a new one, moves left or right, and changes its state. Despite its simplicity, a Turing machine can simulate any step-by-step computation, helping us understand the basics of what computers can and cannot do.

To understand Turing Machine, first you need to know what is 'machine' in Automata.
Automata means "self-acting" Automata is the plural of automation.

Now, consider a black-box in your mind....A magical box....You throw some things into that black-box(input) and that black-box gives you something in return (output).....

Now consider this:

Input---->[black-box]---->Output

orange ----> [juicer/blender] ----> Orange juice

data ----> [computer] ----> information

Here data means "raw facts"
information means "processed raw facts" or "processed data"


Doesn't it sound like a Mathematical function? Even if you don't know anything about them, just see below and you would understand.

y = f(x)

x = input

f = black-box

y = output


x -------------> f---------------> y
[Input] -------------> [black box] ---------------> [output]

Computer ---> something which process data and converts it into information
Data ---> Raw Facts
Information ---> Processed Raw Facts.....
Input ---> Output

computable or not computable ----> "data(raw facts)" to be processed into "information" or not

So, what is computable? This is the fundamental question of CS and one of the most fundamental questions of CS if not the most fundamental...

Now based on these assumptions, we can say that
automata--->abstract machine

Now assume sequences of symbols(discrete) selected from a finite set I of input signals. Namely, set I is the set {x1, x,2, x3... xk} where k is the number of inputs

{x1, x2, x3,.....xk} ------> [black-box] ----------> {y1, y2, y3,....ym}

I --------------> [black-box] --------------> Z

Inputs -----------------------> [black-box]---------------------------> Outputs

Assume Outputs to be sequences of symbols selected from a finite set Z. Set Z is the set {y1, y2, y3 ... ym} where m is the number of outputs.

Discrete is something which is not continuous.....Or something which is continuous but could be cut into parts....For example, the set of real numbers is continuous, including fractions, decimals, and irrational numbers. There are infinite possibilities between any two real numbers. The set of natural numbers (1, 2, 3, ...) is discrete, as there are distinct, separate values with no fractions or decimals.

Now we come to 'state'

{inputs} --------------> [black-box] <states>--------------> {outputs}

State is a resting place while the abstract machine reads more input, if more input is available, states are named.....and named uniquely so we can make a track record.

Imagine 'state' to be a circle with something written inside....I can't make a circle here in the text, so pretend something which is written in a circle is a state....

If it's still sounding confusing, imagine a traffic light. States would be red, green, yellow. It's like a phase. In red state or phase of the traffic signals, cars would stop. To go from one state to another state, a transition is required.

We can make a set of states and we could name it Q.........and we can name our first state to be q0

Point to be noted here is that so far, everything is finite up to this point.

An automaton is called finite if it consists of a finite number of states and functions with finite strings of input and output.

Now what Alan Turner did is he proposed first "infinite" (or unbounded) machine....

A Turing machine has the set [Q, Σ, Γ, δ, q0, B, F] where:

Q = finite set of states, of which one state q0 is the initial state

Γ = finite set of allowable tape symbols

Σ = a subset of Γ not including B, is the set of input symbols

δ = the next move function, a mapping function from Q x Γ to Q x Γ x {L,R}, where L and R denote the directions left and right respectively

q0 = in set Q as the start state

B = a symbol of Γ, as the blank

F ⊆ Q the set of final states

The major difference between a Turing machine and finite automata lies in the fact that the Turing machine is capable of changing symbols on its tape and simulating computer execution and storage. For this reason, it can be said that the Turing Machine has the power to model all computations.

Transition: A way for a machine to go from one state to another given a symbol from the input. Graphically, a transition is represented as an arrow pointing from one state to another, labelled with the symbol or symbols that it can read in order to move the machine from the state at the tail of the arrow to the tip of the arrow. Transitions may even point back to the same state that they came from, which is called a loop.

Tape: Imagine a tape to be an abstract (imaginary) strip of infinite length divided into cells on which input is given.

Every cell includes a symbol from a certain finite alphabet.

Alphabet: Alphabet is one of the ways to represent a data (raw facts)

For example, the computer on which you and I are on work with 1s and 0s. So, the alphabet is just 0 and 1.

But it doesn't mean every alphabet would be 0 and 1. An alphabet in this context just means a set of symbols.

String: A string is a collection of symbols.....

Input : A string fed to a machine which the machine will determine whether it is part of the language that the machine was designed for. The string must only be made up of symbols from the machine's alphabet (e.g. it doesn't make sense to feed a 3 to a machine that only reads 0's and 1's). An input is read by a machine in a forward fashion, one symbol at a time.

Head: It's an entity which reads input (Imagine it like a spotlight on the cell of the tape)

State Register: It stores the state of the Turing machine So we can say that Turing Machine needs storage (I would come back to this later)

After reading an input symbol, it is replaced with another symbol, its internal state is changed, and it moves from one cell to the right or left. If the TM reaches the final state, the input string is accepted, otherwise rejected.

String is a finite sequence of symbols taken from ∑

This will all get cleared with example.

For example, a sting 'aadbcaad' is a valid string (squence of symbols) on the alphabet set ∑ = {a, b, c, d}

another example: a string '00100111' is valid string (sequence of symbols) on the alphabet set ∑ = {0,1}

Imagine a table lamp head which it throws a spotlight on a one of the cells of an infinite tape....

If the machine is in state 1 then an A moves it to state 2 and a B moves it to state 3.

When a symbol, a character from some alphabet, let's say, is input to FSM, it changes state in such a way that the next state depends only on the current state and the input symbol.

So, for example, state 3 needs to have input A to go to some other state (which is state 1 in this case)....

A Turing Machine is a FSM that has::
Infinite tape
reads an input on the tape (which is used as an input to the finite state machine)
This takes the input symbol and according to it and the current state does three things:
---it prints something on the tape
---moves the tape right or left by one cell
---changes to a new state

The simple idea of Turing Machine having infinite tape makes it different from FSM..A Turing machine can also perform a special action – it can stop or halt...Turing machine is said to recognise a sequence of symbols written on the tape if it is started on the tape and halts in a special state called a final state.

For example, a finite state machine can recognise a sequence that has three As followed by three Bs e.g. AAABBB and so can a Turing machine. But:

Only TM can recognise a sequence that has an arbitrary number of As followed by the same number of Bs....In other words, a TM is powerful than a FSM because it can count and for that:

To count that number of Bs equal to number of As, you have to make an equivalence in the form of counting....or matching them up.....and there can be infinite number of As (on the tape because it is infinite) before Bs can occur, so you need unlimited storage to count the inputs to establish equivalence and where is this unlimited storage? It's the very infinite tape on which the head writes! In other words, FSM doesn't have unlimited storage because it does not have infinite tape....And for equivalence, recall the ancient 50,000 years ago concept of more vs less.

The previous set which I explained it was definition, this is one of the examples:

Turing machine TM = [Q, Σ, Γ, δ, q0, B, F]

Q = {q0, q1, q2, qf} [Set of States]
Γ= {a, b} [Tape Alphabet]
∑ = {1} [Input Alphabet]
q0 = {q0} [Initial State]
B = blank symbol
F = {qf } [Final State]

δ, (transitions is given as)

Γ Current State q0 Current State q1 Current State q2
a 1Rq1 1Lq0 1Lqf
b 1Lq2 1Rq1 1Rqf

"1Rq1" can be interpreted as follows: "If the machine is in state q0, and it reads the symbol '1' from the tape, it will move the tape head one position to the right(R) and transition to state q1." Similarly, other transition rules are interpreted according to their nature until we reach qf (final state). A visual diagram would have made it much easier instead of a table, but since it's a text post, my hands are tied.

This simple setup can mimic the logic of any computer program, showing the fundamental principles of computation. It's like a basic, universal model of how computers work. However, it's to be noted that it's a theoretical model.

Coming towards AI, these foundational aspects are indirectly related to AI. Turing's work, particularly his concept of the Turing machine, provided a theoretical foundation for computation. This theoretical framework influenced the development of algorithms and the understanding of what it means to solve a problem algorithmically. AI algorithms, being a subset of computational algorithms, are built on these principles.

Turing's work contributed to the understanding of what problems are computable and what problems are undecidable. This has implications for AI, as it deals with complex problem-solving. The concept of undecidability, such as the halting problem, highlights the limitations of building a general algorithm that can predict the behaviour of any other algorithm.

The foundation of AI spans various disciplines, including computer science, mathematics, cognitive science, and philosophy. I would not try to define AI because its definition varies. No one can give a concrete answer to that. Everyone would have a different opinion about it.

When we say AI in arts, it's such broad term that various technologies come into play. It's a very generic term, but at the same time, very broad. For the sake of this post, and to be specific, people consider generative text to visual to be AI in arts (I am talking about visual art). A generic term for it would be 'generative AI'.

I would try to explain this in much simpler way.

Generative AI, particularly in the context of transforming text into images, often involves techniques such as Generative Adversarial Networks (GANs) introduced by Ian Goodfellow and his colleagues., and other generative models. GANs are a class of machine learning models that consist of two neural networks: a generator and a discriminator. These networks are trained simultaneously through adversarial training.

Before going further, I would first simplify the terms of the previous paragraph. Machine learning is a subset of artificial intelligence (AI) that focuses on developing algorithms and models that enable computers to learn from data. Instead of being explicitly programmed, these algorithms learn patterns and make predictions or decisions. A neural network is a computational model inspired by the structure and functioning of the human brain. It consists of interconnected nodes (neurons) organized into layers. Neural networks are used in machine learning to learn patterns from data. A machine learning model is a mathematical representation of a real-world process or system. It is trained on data to make predictions or decisions without being explicitly programmed for each task. The generator is part of a Generative Adversarial Network (GAN). Its role is to create new data (e.g., images) that resembles the training data. It takes random noise as input and generates samples. The discriminator serves as a "judge." Its primary function is to differentiate between real data, obtained from the training set, and synthetic data generated by the generator. Over the course of training, the discriminator undergoes improvement as it becomes more adept at distinguishing between real and artificially created samples.

As the terms have been explained, let's go further. In the context of transforming text into images, the process involves training a GAN on a dataset of paired text and corresponding images. The generator is tasked with creating realistic images based on textual descriptions, while the discriminator's role is to distinguish between real images from the dataset and those generated by the generator.

During training, the generator and discriminator are engaged in a game. The generator tries to create images that are indistinguishable from real images, while the discriminator tries to get better at telling real images apart from generated ones. This adversarial process continues until the generator produces high-quality images that are difficult for the discriminator to distinguish. It's like Phoenix Wright (attorney) winning the case where judge can't even judge anymore.

Conditional GANs extend the concept by introducing additional information, such as class labels or textual descriptions, to guide the generation process. In the case of text-to-image generation, the generator is conditioned on textual input to produce images that match the provided description. (I would come back to this 'match the provided description' in a funny way).

While text-to-image generation has seen progress, it still faces challenges, such as generating diverse and contextually accurate images based on textual input. Ensuring that generated images align with the intended meaning of the text remains an ongoing research area.

The word 'context' here is very important. I would also come to this later.

To have an example of the concept, a textual description is provided as input. For example: "A banana on a glass table." The generator uses the input text to produce an image that matches the description. The generated image might depict a banana on a glass table. The generator's output is compared with real images, and the discriminator provides feedback. The generator adjusts its parameters to improve the quality of generated images. The process is repeated iteratively until the generator produces visuals that closely align with the input text.

As we just tried to grasp a little bit of the concept of how it works on a rudimentary and mundane level, we can now discuss the implications and relationship of it with respect to art.

Before coming towards that, there are three import things in this process I would like to discuss. Data, context, and methodology (training). To start, I pose a question that where does data come from? It's a deep question. And even before this question, what actually 'data' is? I already defined data in the post. But let's break the conventional shackles and think beyond that. Let's combine the questions, "What is data and where does it come from?"

It's such a deep question that no written book would ever be enough. Sometimes, I feel that a human mind can ever be able to comprehend the question at all. Definitely one of my most favourite questions of all time. I would only give my philosophical impressions on that.

When we say 'data', what do we mean by it? Let me become a devil for an instance and say, 'eliminate the humans out of it'. Before us, data was there and if humanity would not exist, data would still be there. But when humans are added into the equation, then data itself would not have any meaning unless we deem it worthy enough. It then becomes 'your data', 'my data', 'their data', 'our data'. It either becomes a selfish capitalist or a obnoxious communist. In either case, it all comes back to more data vs less data. The ancient 50,000 years ago of more vs less. One thing is certain. Data flows in the same way as time does.

As time flows, knowledge gets transferred. Some gets lost. Generations after generation, it flows just as time. Those ontological philosophers who talk about the existence and nature of data—whether it has an independent existence or if it is a human construct still can't answer that. I believe the former. I believe it to have an independent existence. Some call it to be the raw material of knowledge. . Constructivists argue that knowledge is actively constructed by individuals based on their experiences and interpretations. However, what data is, it's a mystery. It's something that we are unable to fathom. The direct association of where it comes from is another mystery. A mystery so captivating. If it comes from somewhere, where does it reside? Can we store it permanently? Or can it never be stored, but only transferred? That blurs the line between storage and transference.

As datasets being used in the training, the art world is in fumes. They have a notion that that their work was stolen or used without their permission. This is a complex issue that is related to legal domain. I won't touch this topic at all. Ethics are also subjective. However, I can share some musings. From my lenses, if the generator companies are commercially utilizing the generators (which they are ofc), then it's their ethical and moral bankruptcy. It's going to bite them in long-term. If they are doing private research without the commercialization, even then, it's unethical because they should have engaged and involved the art world to build a legal framework first. However, the art world that is utilizing the computer in their works also need to know that just a single point clicks of their mouse, there is insane amount of people's work which came generation after generation. Hatred is not going to benefit anyone. It's insanity everywhere. That's why context is paramount.

One thing is certain. Regardless of stolen works in the datasets being present or not, the technology would have matured and would have come sooner or later. Pushing it back is just time delay. The argument that is being made that music industry data wasn't infringed because they had much better protection system, regardless of that, the technology still exists. Try Suno AI and see how well it's producing songs. Much better than the generative images in my opinion if by visual art world's woes. Whether the infringed work present in the dataset or not, the end result would be more or less the same. If not now, then in future.

Then, stolen content just doesn't work that it's scrapped from any official portal. The web crawlers are enough. The art world needs to look at the way how web works. Any data that becomes a part of databed for testing does not need to come from official portals. The web is not a safe house where in real life, one rings a door bell to seek permission. The works become a part of unintended consequences. It's the nature of the beast of big data.

In the future, there would be very personalized generative AIs that you might want to train them based on your gallery work visuals and after sufficient fine tuning, they would be your personal art pets. They would now mimic your personality and style. You can then order them by giving your ideas to them as an art director and they would speedily produce your feasible ideas. I personally think this more ethical way to utilize personalization in AI art. Others might disagree.

I would also not talk about the job or replacement aspect of this. I want to point out that art world feels devastated and insecure. I have seen people crying and are scared. Not only from art world, but from other domains as well. So many people have lost motivation to pick the brush and paint. Some say that those who never drew a single line in their lives would outperform them. I would present my thoughts about this aspect next.

I think the notion from art world that anyone could outperform them needs to be dealt with philosophical perspective. It all comes to personal take on your own art. The basic questions of why do you do and for what? From my perspective, it's like there is a washing machine where tech people have stolen and thrown the art people's different clothes of different materials and colours, and the washing machine has returned the clothes to art people where it's a huge mixture of someone's polka dots attached to someone else's fibres and other intermingled mess of materials and colours. And people are arguing over it. Guns didn't stop people from going into dojos. Cars and rollercoasters didn't stop people from walking. Fast food hasn't stopped gyms. There will always be time and place for your hand made art. You would only give in to despair if you really would let it.

If you honestly think that prompting is that scary that anyone can trample over you, just let me give you one little example. I am going to give you all a prompt and challenge any generator out there to make it. Here it is:

"Create an image of my childhood class 3 science class that no longer exists now where my best friends assembled, and I tried to draw superman on chairs". What happened? It's a failure because no AI not now, nor in the future would be able to do that because that scenario doesn't exist in the real world because that school building is extinct and got replaced by something else now. If you are skilled artist, you would have that school in your mind, and only you are able to paint that scenario, not AI.

This is just one little example. What caused the defeat of AI prompt wasn't the artwork nor the methodology, it was the idea even though it was specific. The prompt was language-wise true and fulfilled the criteria, but it was relative truth. It wasn't the full truth because even though the artwork got generated, it wasn't true for you. It got failed because of the idea itself and that idea came from your mind world. The idea can work in your art and can be brought into fruition, in your truth world, but in AI's world, it would never. You can even write a prompt book describing each aspect of the idea but would never get the result because that scenario exists in only your mind and only you can bring it out visually. AI does not know that class or school unless I give it a visual reference which also does not exist.

I opted for a personalized touch in my concept and AI failed miserably. Let's call this kind of concept 'personalization of memories'. These memories could be of past, present or future, real or hypothetical. It's just one example. This is why I say Millennial artists should never lose hope. They are born in a crossroad of century when social media pictures didn't exist for data acquiring.

The time has come where artists now need to revive those lost personalized ideas that never got the importance and value. Make them valuable now. And there are lots of them. Unlimited...Methodology is the least thing artists should be concerned about. It's like Wonderweiss sealed the flames and Yamamoto called him so naive that Wonderwiess thought that his zanpakutou (endpoint) should be feared the most and not his martial arts (methodology). Every artist finds the joy by moving their hands, that's a given, but the real way to stand yourself out from the AI are the ideas themselves.

I personally lean towards those group of people who value journey to be more important than the destination. All my trips I did, I remember the journey more than the endpoint. The fun is truly in the route. AI art values endpoint more by eliminating the joy of the journey. The path is personal, endpoint is public. No matter how much the endpoint of AI art visual looks polished, the joy or journey of that methodology can only be experienced by you. Instead of feeling belittled, respect your methodology and craft. Be really proud of it. These AI generators might give you the power of being an art director to some extent, but it's the sanctity of idea that would win which only comes from your crafting. The sanctity is subjective, but it should be you who would determine that. Personalization with your story and personal context is your weapon against AI art. It's the weapon of love for your craft and methodology. AI fears that, truly.

Lastly, I don't share my original works here, but for the sanctity of this post, I am going to share some highlights of one of my original works. I won't be sharing the high-quality image because in my personal site, I lock the image accessibility which can only be viewed there. Secondly, it wasn't a BLEACH work also.

About this project, the idea is something that has a personal touch to it and can't be generated by AI. The idea is of my school friend whom I had a great friendship with. I don't know where she is right now, but she always existed in my mental world. I first tried to make her a zombie, but later, I changed my mind. The whole art is the pure manifestation of my mental world for her. This idea can never be generated with AI because AI never knows that person in the state which she was in my mental world. AI might have her data if she uploaded an image of her publicly but at that time of school, there was no camera or social media. Plus, I don't have any picture of her either. It's only my memories that preserved the dreamy visual of her. Secondly, her state of being changed in my mind as I made her a fairy. No amount of prompts would be able to describe the visual personification of it and not possible with AI because it would never know her state.
 
Last edited:
Top