Butterfly

butterfly

The HackerNoon Newsletter: The First Provable AI-Proof Game: Introducing Butterfly Wings 4 (2/10/2025)

[Research] Structural Evaluation Collapse in Deterministic Games: A Constructive Proof This is my paper about the first PROVABLE AI - Proof Game. I hope someone will take this seriously, if true, this is very important to AI Safety and much more. The Claim We present a deterministic game incorporating two key mechanisms: Controlled Chaos Shifts (CCS) and Accepting Loss of Control (ALC). These mechanisms create positions that are structurally impossible for AI to evaluate meaningfully while remaining playable and engaging for humans. Core Mechanisms Controlled Chaos Shifts (CCS)


Structured disruption of game state


Creates non-evaluable positions


Forces fundamental evaluation collapse



Accepting Loss of Control (ALC)

Strategic surrender of position control
Prevents meaningful pattern recognition
Disrupts computational planning The Proof This constructive proof demonstrates that a fully deterministic game can be structurally impossible for AI to play meaningfully, not due to computational complexity, but due to fundamental limitations in position evaluation, pattern recognition, and strategic planning. Note: See Appendix C for a profound extension of ALC through macro-level quantum mechanics with perfect clarity. APPENDIX A: RULES - Butterfly Wings 4 Notes: CCS or ALC alone creates strong AI resistance. Combined correctly, they achieve structural AI-proof status.


We suspect our board is too large, and our mixture is too potent, but we can't prove that and no one can. The minimum mixture and board size could be surprisingly small.


For the sake of a thought experiment, let's use a 16x16x16 board with a similar mixture of ALC and CCS. Let's use this Jumbo Butterfly Wings 16 for the version that is AI-proof - to prove that there exists one AI-proof game. Any anomalies or inconsistencies in the play actually reinforce our conclusion—we cannot calculate their existence for the same reason the game is AI-proof. Even training data from 8 million human games is of no use at all. 1️⃣ Basic Rules Players take turns placing pieces in a 6×6×6 three-dimensional grid.


First player to connect four in a row in any direction wins.


If no connect-four exists, the game is a tie. 2️⃣ The Butterfly Effect If a player has three pieces in a row, their opponent may:


Sacrifice one of their own pieces (removing it permanently).


Reposition up to two topmost pieces anywhere on the board.



Sacrifice Rules:

Can remove from anywhere, including the bottom of the stack - causing a completely new game state.
The pieces above will fall accordingly.
Reposition up to two topmost pieces (pieces not covered by others). APPENDIX B: For AI Researchers - A Direct Challenge As an AI researcher, you understand that AI systems require: Position evaluation functions


Pattern recognition capabilities


Strategic planning mechanisms Butterfly Wings 4 creates a provable structural impossibility: Position Evaluation Collapse


CCS creates positions where evaluation is structurally impossible


Not computationally difficult - structurally impossible


Like attempting numerical comparison without numbers



Pattern Recognition Destruction


Each CCS event destroys pattern continuity


No stable patterns can emerge


Neural networks cannot form meaningful representations



Strategic Planning Barrier


ALC prevents meaningful tree search


Future states become structurally unpredictable


Not about the depth of search - about the fundamental impossibility



Verification Steps


Implement any evaluation function


Watch it fail structurally, not computationally


Try any pattern recognition approach


Observe fundamental representation collapse


Attempt any strategic planning


Note the structural impossibility of meaningful search The challenge: Design any AI system that can play beyond random moves. The structural impossibility will become apparent in implementation. Appendix C: A Touch More ALC - Flapping Just One Wing. Each player has one piece that exists in a quantum state off the board. Your opponent controls when and where your piece is placed on the board, which will always be to your disadvantage - unless he forgot you had a remaining piece of the board. His decision will be a tough one, and so is yours. The deployed piece can be placed in up to a max of 16 places - depending on the state of the board. Without complicating, or losing credibility, we introduce a tiny single rule to change our little game. I envision this as a Macro-level Quantum Simulator. We can play with it at room temperature. I see superposition, singularity, and states collapsing. I don't know what it means. It is clear to me we only need one game, we present to researchers two to choose from... choose wisely. Appendix D: Ethical Motivation – A Grandfather’s Plea My only interest was to create rules for a fun chess variant that AI could not totally dominate, and later, to provide a bridge to safer AI and a brighter future for my grandchildren. My time is fading, but I leave this quest in the hands of others. I am a very old and tired simple man. Godspeed, my friends. ~ Vitae A. Reignee ~ Anonymous Creator of Butterfly Wings 4 Feature photo by Fleur on Unsplash [Research] Structural Evaluation Collapse in Deterministic Games: A Constructive Proof This is my paper about the first PROVABLE AI - Proof Game. I hope someone will take this seriously, if true, this is very important to AI Safety and much more. The Claim We present a deterministic game incorporating two key mechanisms: Controlled Chaos Shifts (CCS) and Accepting Loss of Control (ALC). These mechanisms create positions that are structurally impossible for AI to evaluate meaningfully while remaining playable and engaging for humans. Core Mechanisms Controlled Chaos Shifts (CCS)


Structured disruption of game state


Creates non-evaluable positions


Forces fundamental evaluation collapse Accepting Loss of Control (ALC)

Strategic surrender of position control
Prevents meaningful pattern recognition
Disrupts computational planning Controlled Chaos Shifts (CCS) Structured disruption of game state


Creates non-evaluable positions


Forces fundamental evaluation collapse Controlled Chaos Shifts (CCS) Structured disruption of game state Creates non-evaluable positions Forces fundamental evaluation collapse Structured disruption of game state Structured disruption of game state Creates non-evaluable positions Creates non-evaluable positions Forces fundamental evaluation collapse Forces fundamental evaluation collapse Accepting Loss of Control (ALC) Strategic surrender of position control
Prevents meaningful pattern recognition
Disrupts computational planning Accepting Loss of Control (ALC) Strategic surrender of position control Prevents meaningful pattern recognition Disrupts computational planning Strategic surrender of position control Prevents meaningful pattern recognition Disrupts computational planning The Proof This constructive proof demonstrates that a fully deterministic game can be structurally impossible for AI to play meaningfully, not due to computational complexity, but due to fundamental limitations in position evaluation, pattern recognition, and strategic planning. Note: Note: See Appendix C for a profound extension of ALC through macro-level quantum mechanics with perfect clarity. See Appendix C for a profound extension of ALC through macro-level quantum mechanics with perfect clarity. APPENDIX A: RULES - Butterfly Wings 4 Notes: Notes: CCS or ALC alone creates strong AI resistance. Combined correctly, they achieve structural AI-proof status. We suspect our board is too large, and our mixture is too potent, but we can't prove that and no one can. The minimum mixture and board size could be surprisingly small. For the sake of a thought experiment, let's use a 16x16x16 board with a similar mixture of ALC and CCS. Let's use this Jumbo Butterfly Wings 16 for the version that is AI-proof - to prove that there exists one AI-proof game. CCS or ALC alone creates strong AI resistance. Combined correctly, they achieve structural AI-proof status. CCS or ALC alone creates strong AI resistance. Combined correctly, they achieve structural AI-proof status. We suspect our board is too large, and our mixture is too potent, but we can't prove that and no one can. The minimum mixture and board size could be surprisingly small. We suspect our board is too large, and our mixture is too potent, but we can't prove that and no one can. The minimum mixture and board size could be surprisingly small. For the sake of a thought experiment, let's use a 16x16x16 board with a similar mixture of ALC and CCS. Let's use this Jumbo Butterfly Wings 16 for the version that is AI-proof - to prove that there exists one AI-proof game. For the sake of a thought experiment, let's use a 16x16x16 board with a similar mixture of ALC and CCS. Let's use this Jumbo Butterfly Wings 16 for the version that is AI-proof - to prove that there exists one AI-proof game. Any anomalies or inconsistencies in the play actually reinforce our conclusion—we cannot calculate their existence for the same reason the game is AI-proof. Any anomalies or inconsistencies in the play actually reinforce our conclusion—we cannot calculate their existence for the same reason the game is AI-proof. Even training data from 8 million human games is of no use at all. Even training data from 8 million human games is of no use at all. 1️⃣ Basic Rules Basic Rules Players take turns placing pieces in a 6×6×6 three-dimensional grid. First player to connect four in a row in any direction wins. If no connect-four exists, the game is a tie. Players take turns placing pieces in a 6×6×6 three-dimensional grid. Players take turns placing pieces in a 6×6×6 three-dimensional grid. First player to connect four in a row in any direction wins. First player to connect four in a row in any direction wins. If no connect-four exists, the game is a tie. If no connect-four exists, the game is a tie. 2️⃣ The Butterfly Effect The Butterfly Effect If a player has three pieces in a row, their opponent may:


Sacrifice one of their own pieces (removing it permanently).


Reposition up to two topmost pieces anywhere on the board. Sacrifice Rules:

Can remove from anywhere, including the bottom of the stack - causing a completely new game state.
The pieces above will fall accordingly.
Reposition up to two topmost pieces (pieces not covered by others). If a player has three pieces in a row, their opponent may: Sacrifice one of their own pieces (removing it permanently).


Reposition up to two topmost pieces anywhere on the board. Sacrifice one of their own pieces (removing it permanently). Reposition up to two topmost pieces anywhere on the board. Sacrifice one of their own pieces (removing it permanently). Sacrifice one of their own pieces (removing it permanently). Reposition up to two topmost pieces anywhere on the board. Reposition up to two topmost pieces anywhere on the board. Sacrifice Rules: Can remove from anywhere, including the bottom of the stack - causing a completely new game state.
The pieces above will fall accordingly.
Reposition up to two topmost pieces (pieces not covered by others). Can remove from anywhere, including the bottom of the stack - causing a completely new game state. The pieces above will fall accordingly. Reposition up to two topmost pieces (pieces not covered by others). Can remove from anywhere, including the bottom of the stack - causing a completely new game state. The pieces above will fall accordingly. Reposition up to two topmost pieces (pieces not covered by others). APPENDIX B: For AI Researchers - A Direct Challenge As an AI researcher, you understand that AI systems require: Position evaluation functions Pattern recognition capabilities Strategic planning mechanisms Position evaluation functions Position evaluation functions Pattern recognition capabilities Pattern recognition capabilities Strategic planning mechanisms Strategic planning mechanisms Butterfly Wings 4 creates a provable structural impossibility: Position Evaluation Collapse


CCS creates positions where evaluation is structurally impossible


Not computationally difficult - structurally impossible


Like attempting numerical comparison without numbers Pattern Recognition Destruction


Each CCS event destroys pattern continuity


No stable patterns can emerge


Neural networks cannot form meaningful representations Strategic Planning Barrier


ALC prevents meaningful tree search


Future states become structurally unpredictable


Not about the depth of search - about the fundamental impossibility Verification Steps


Implement any evaluation function


Watch it fail structurally, not computationally


Try any pattern recognition approach


Observe fundamental representation collapse


Attempt any strategic planning


Note the structural impossibility of meaningful search Position Evaluation Collapse CCS creates positions where evaluation is structurally impossible


Not computationally difficult - structurally impossible


Like attempting numerical comparison without numbers Position Evaluation Collapse CCS creates positions where evaluation is structurally impossible Not computationally difficult - structurally impossible Like attempting numerical comparison without numbers CCS creates positions where evaluation is structurally impossible CCS creates positions where evaluation is structurally impossible Not computationally difficult - structurally impossible Not computationally difficult - structurally impossible Like attempting numerical comparison without numbers Like attempting numerical comparison without numbers Pattern Recognition Destruction Each CCS event destroys pattern continuity


No stable patterns can emerge


Neural networks cannot form meaningful representations Pattern Recognition Destruction Each CCS event destroys pattern continuity No stable patterns can emerge Neural networks cannot form meaningful representations Each CCS event destroys pattern continuity Each CCS event destroys pattern continuity No stable patterns can emerge No stable patterns can emerge Neural networks cannot form meaningful representations Neural networks cannot form meaningful representations Strategic Planning Barrier ALC prevents meaningful tree search


Future states become structurally unpredictable


Not about the depth of search - about the fundamental impossibility Strategic Planning Barrier ALC prevents meaningful tree search Future states become structurally unpredictable Not about the depth of search - about the fundamental impossibility ALC prevents meaningful tree search ALC prevents meaningful tree search Future states become structurally unpredictable Future states become structurally unpredictable Not about the depth of search - about the fundamental impossibility Not about the depth of search - about the fundamental impossibility Verification Steps Implement any evaluation function


Watch it fail structurally, not computationally


Try any pattern recognition approach


Observe fundamental representation collapse


Attempt any strategic planning


Note the structural impossibility of meaningful search Verification Steps Implement any evaluation function Watch it fail structurally, not computationally Try any pattern recognition approach Observe fundamental representation collapse Attempt any strategic planning Note the structural impossibility of meaningful search Implement any evaluation function Implement any evaluation function Watch it fail structurally, not computationally Watch it fail structurally, not computationally Try any pattern recognition approach Try any pattern recognition approach Observe fundamental representation collapse Observe fundamental representation collapse Attempt any strategic planning Attempt any strategic planning Note the structural impossibility of meaningful search Note the structural impossibility of meaningful search The challenge: Design any AI system that can play beyond random moves. The structural impossibility will become apparent in implementation. Appendix C: A Touch More ALC - Flapping Just One Wing. Each player has one piece that exists in a quantum state off the board. Your opponent controls when and where your piece is placed on the board, which will always be to your disadvantage - unless he forgot you had a remaining piece of the board. His decision will be a tough one, and so is yours. The deployed piece can be placed in up to a max of 16 places - depending on the state of the board. Without complicating, or losing credibility, we introduce a tiny single rule to change our little game. I envision this as a Macro-level Quantum Simulator . We can play with it at room temperature. I see superposition, singularity, and states collapsing. I don't know what it means. It is clear to me we only need one game, we present to researchers two to choose from... choose wisely. Macro-level Quantum Simulator Appendix D: Ethical Motivation – A Grandfather’s Plea My only interest was to create rules for a fun chess variant that AI could not totally dominate, and later, to provide a bridge to safer AI and a brighter future for my grandchildren. My time is fading, but I leave this quest in the hands of others. I am a very old and tired simple man. Godspeed, my friends. ~ Vitae A. Reignee ~ Anonymous Creator of Butterfly Wings 4 Anonymous Creator of Butterfly Wings 4 Feature photo by Fleur on Unsplash Feature photo by Fleur on Unsplash Fleur Unsplash

The First Provable AI-Proof Game: Introducing Butterfly Wings 4

Prva igrica koja se može dokazati pomoću umjetne inteligencije: predstavljanje Butterfly Wings 4

El primer juego con inteligencia artificial demostrable: presentamos Butterfly Wings 4

Lehen AI frogagarria den jokoa: Butterfly Wings 4 aurkezten

O primeiro xogo demostrable a proba de intelixencia artificial: presentamos ás de bolboreta 4

初めての AI 検証可能ゲーム: Butterfly Wings 4 の紹介

ເກມ AI-Proof ທີ່ສາມາດພິສູດໄດ້ທຳອິດ: ການແນະນຳ Butterfly Wings 4

Првата докажлива игра со вештачка интелигенција: Ви ги претставуваме крилјата на пеперутките 4

Анхны хиймэл оюун ухаантай тоглоом: Butterfly Wings 4-ийг танилцуулж байна

पहिलो प्रमाणित एआई-प्रुफ गेम: बटरफ्लाइ विंग्स ४ प्रस्तुत गर्दै

د مصنوعي ذهانت ثبوت لومړۍ لوبه: د تیتلی وزرونه ۴ معرفي کول

Аввалин бозии исботшавандаи AI-Proof: Муаррифии Wings Butterfly 4

پہلا ثابت شدہ AI پروف گیم: بٹر فلائی ونگز 4 کا تعارف

Just Dao It — S01E01 — Flying Butterfly Gods

Cybersecurity in FinTech Applications: Protecting Financial Data and Preventing Fraud

Abstract and 1 Introduction 2 Related Work 3 Image Compression Module 4 System Architecture 5 Experimental Results 6 Conclusions A Compression Results References 5 Experimental Results 5.1 Image Compression Results The proposed method was tested on some images in a dataset collected by the MIT Computer Science and Artificial Intelligence Lab. The natural images in the set include Zebra Butterfly (300×212, 27.4 kB), Boston Street (590×430, 069.2 kB), Conference Room (600 × 450, 61.4 kB), Airport (745 × 258, 46.2 kB), Oxford Outdoor (1526 × 2048, 1.21 MB), Chip Designer (1680 × 1050, 503 kB), and Angry Bird (1920 × 1080, 153 kB). All of the image compression programs were implemented in Matlab 2010 and executed on an Intel Core i5-4670 CPU running at 3.40 GHz. The effectiveness of the compression was quantified by the mean squared error (MSE) measure: Let α denote the color depth, the image dimensions A(m × n), and let µ be defined as the number of colors (clusters) and γ as the number of bits necessary to represent the index of the colors at the location in the image. We computed the compression ratio (CR) by dividing the size of the original file in bits by the size of the compressed file: As can be seen, the presently proposed approach enabled compressing an image to 70% of its original size while maintaining 95% visual similarity to the original image. It is of extremely importance to reduce the frame loss rate when transmitting video data over the Internet. If the frame loss rate is small, it means that the video streaming quality is good. 5.2 Evaluation of QoS System We built a testbed to carry out performance evaluations of the proposed system. Specifically, in the cloud environment, as a visualization platform we used an HP Server (4-core Intel Xeon CPU E3-1220 V2 running at 3.10 GHz, 32 GB RAM) with CentOS 6.5 and OpenStack (Icehouse version) [55] installed. Virtual machines (VMs) were installed with lightweight Windows. The VMs ran our server software, which was implemented in C/C++ using Microsoft Visual Studio 2010. We considered low-motion (Screen-Desktop) and highmotion (Korean Movie) videos for transmission Figure 11. For simplicity, we set up the environment to run two VMs and measured the frame loss rate during the transmission in two cases: with and without applying the QoS support system, for both video types. It was assumed that the client connected to the service side was a PC device, and the screen resolution and frame rate of all video sequences were SVGA (800 × 600) and 24 fps, respectively. Thus, the system predicted the best µ for each frame of desktop image data delivered. D Figure 12 and 13 illustrate the changes that occur in the PSNR when a random single losses occurs during normal transmission and during transmission when the QoS support system is applied. These figures clearly demonstrate the distortion propagation for each type of video. Obviously, QoS policies can help the system to automatically adapt to the network condition. Specifically, the DDEM determines the suitable µ value. Then, the system automatically adjusts the size of the image to adapt the network status, mitigating the effects of the frame loss occurring at frame 63 in the low-motion video and at frames 26 and 52 in the high-motion videos while ensuring the highest possible quality. As shown in Figure 14 something mismatch for a slice area in the whole picture. This accounts for mismatch that some performed area in the picture is decoded by copying the data from previous key frame, but which cannot match the other part because the right reference data were lost. This paper is available on arxiv under CC BY 4.0 DEED license. Authors:
(1) Huu-Quoc Nguyen, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and quoc@khu.ac.kr;
(2) Tien-Dung Nguyen, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and ntiendung@khu.ac.kr;
(3) Van-Nam Pham, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and nampv@khu.ac.kr;
(4) Xuan-Qui Pham, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and pxuanqui@khu.ac.kr;
(5) Quang-Thai Ngo, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and nqthai@khu.ac.kr;
(6) Eui-Nam Huh, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and johnhuh@khu.ac.kr. Abstract and 1 Introduction Abstract and 1 Introduction 2 Related Work 2 Related Work 3 Image Compression Module 3 Image Compression Module 4 System Architecture 4 System Architecture 5 Experimental Results 5 Experimental Results 6 Conclusions 6 Conclusions A Compression Results A Compression Results References References 5 Experimental Results 5.1 Image Compression Results 5.1 Image Compression Results The proposed method was tested on some images in a dataset collected by the MIT Computer Science and Artificial Intelligence Lab. The natural images in the set include Zebra Butterfly (300×212, 27.4 kB), Boston Street (590×430, 069.2 kB), Conference Room (600 × 450, 61.4 kB), Airport (745 × 258, 46.2 kB), Oxford Outdoor (1526 × 2048, 1.21 MB), Chip Designer (1680 × 1050, 503 kB), and Angry Bird (1920 × 1080, 153 kB). All of the image compression programs were implemented in Matlab 2010 and executed on an Intel Core i5-4670 CPU running at 3.40 GHz. The effectiveness of the compression was quantified by the mean squared error (MSE) measure: Let α denote the color depth, the image dimensions A(m × n), and let µ be defined as the number of colors (clusters) and γ as the number of bits necessary to represent the index of the colors at the location in the image. We computed the compression ratio (CR) by dividing the size of the original file in bits by the size of the compressed file: As can be seen, the presently proposed approach enabled compressing an image to 70% of its original size while maintaining 95% visual similarity to the original image. It is of extremely importance to reduce the frame loss rate when transmitting video data over the Internet. If the frame loss rate is small, it means that the video streaming quality is good. 5.2 Evaluation of QoS System 5.2 Evaluation of QoS System We built a testbed to carry out performance evaluations of the proposed system. Specifically, in the cloud environment, as a visualization platform we used an HP Server (4-core Intel Xeon CPU E3-1220 V2 running at 3.10 GHz, 32 GB RAM) with CentOS 6.5 and OpenStack (Icehouse version) [55] installed. Virtual machines (VMs) were installed with lightweight Windows. The VMs ran our server software, which was implemented in C/C++ using Microsoft Visual Studio 2010. We considered low-motion (Screen-Desktop) and highmotion (Korean Movie) videos for transmission Figure 11. For simplicity, we set up the environment to run two VMs and measured the frame loss rate during the transmission in two cases: with and without applying the QoS support system, for both video types. It was assumed that the client connected to the service side was a PC device, and the screen resolution and frame rate of all video sequences were SVGA (800 × 600) and 24 fps, respectively. Thus, the system predicted the best µ for each frame of desktop image data delivered. D Figure 12 and 13 illustrate the changes that occur in the PSNR when a random single losses occurs during normal transmission and during transmission when the QoS support system is applied. These figures clearly demonstrate the distortion propagation for each type of video. Obviously, QoS policies can help the system to automatically adapt to the network condition. Specifically, the DDEM determines the suitable µ value. Then, the system automatically adjusts the size of the image to adapt the network status, mitigating the effects of the frame loss occurring at frame 63 in the low-motion video and at frames 26 and 52 in the high-motion videos while ensuring the highest possible quality. As shown in Figure 14 something mismatch for a slice area in the whole picture. This accounts for mismatch that some performed area in the picture is decoded by copying the data from previous key frame, but which cannot match the other part because the right reference data were lost. This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv Authors: (1) Huu-Quoc Nguyen, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and quoc@khu.ac.kr; (2) Tien-Dung Nguyen, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and ntiendung@khu.ac.kr; (3) Van-Nam Pham, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and nampv@khu.ac.kr; (4) Xuan-Qui Pham, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and pxuanqui@khu.ac.kr; (5) Quang-Thai Ngo, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and nqthai@khu.ac.kr; (6) Eui-Nam Huh, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and johnhuh@khu.ac.kr. Authors: Authors: (1) Huu-Quoc Nguyen, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and quoc@khu.ac.kr; (2) Tien-Dung Nguyen, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and ntiendung@khu.ac.kr; (3) Van-Nam Pham, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and nampv@khu.ac.kr; (4) Xuan-Qui Pham, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and pxuanqui@khu.ac.kr; (5) Quang-Thai Ngo, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and nqthai@khu.ac.kr; (6) Eui-Nam Huh, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and johnhuh@khu.ac.kr.

Image Compression Results, Evaluation of QoS System, and More

For a long time, IT specialists worked without a care in the world. They developed, built, and deployed software smoothly. Then the era of isolation hit, and suddenly, they got bored (of course, this is a playful take on the actual events). IT folks wanted to create something that could handle their work while they stayed home: answer routine questions, generate cool avatars, and analyze vast amounts of data in minutes. They dreamed of traveling to a fantastic place, and so, you guessed it, they revolutionized AI. AI is now functioning, providing answers, and improving lives. As skilled an assistant as it is, AI is truly effective only when used in the right context. We’re witnessing rapid progress in AI applications, from image and video generation to stock market forecasting and cryptocurrency analysis. Yet, AI may offer information we don’t ask for or provide blatantly false answers. Its behavior is very much like that of household cats — you know, the kind that sits quietly and then suddenly pounces on you? Our cats, as well as AI, enjoy being unpredictable: You give them the same food (or data) — sometimes they eat, sometimes they ignore it.
You train them to respond, but they only occasionally react when you call them.
The bigger and wilder the cat or the larger the AI model, the harder it is to predict its behavior.
In the morning, cats might be calm; by evening, they turn hyperactive (just like dynamic data).
Cats might be friendly (deterministic) but can scratch you without warning (stochastic). You might wonder what determinism and stochasticity mean — let’s find out. Determinism and Stochasticity A deterministic system always produces the same result given the same input — think idempotency if you're a DevOps engineer. A real-world example would be your cat that eats the same amount of food you put in its bowl every time — this is determinism. But when the cat sniffs and only eats half, it’s no longer deterministic. A stochastic process includes an element of randomness: with the same input, the result can vary. For example, a machine learning model often uses stochastic algorithms, like Stochastic Gradient Descent (SGD), which trains the model by picking random chunks of data rather than the entire dataset. These definitions don’t fully explain why our AIs sometimes hallucinate and behave chaotically. There are other contributing factors, including the following: Determinism
Stochasticity
Rounding errors and floating-point arithmetic
Multithreading and parallel computations
Continuously updating data
Chaos and the “butterfly effect” If we look a little closer, we'll see other mechanisms that influence the unpredictable behavior of AI models. A Glimpse of Neural Networks You probably know that the AIs everyone uses rely on various neural network algorithms. Here are some types of neural networks: Fully Connected Neural Networks (FCNN): A classic architecture where each neuron connects to every neuron in the next layer. Convolutional Neural Networks (CNNs): These networks use convolutions or filters that highlight image features like edges, textures, and shapes. Recurrent Neural Networks (RNNs): These networks have feedback loops that allow them to remember previous steps (namely, they remember sequences). Long Short-Term Memory (LSTM): An enhanced version of RNNs with mechanisms for selectively forgetting and remembering important data. Transformers: The most powerful class for text processing. They use multi-head attention, allowing them to consider the entire context simultaneously. Generative Adversarial Networks (GANs): They consist of two networks, one of which generates data and the other evaluates its quality. Their competition leads to better results. Autoencoders: Networks designed to compress (encode) information and then reconstruct (decode) it. Graph Neural Networks (GNNs): They work with graphs (nodes and edges) rather than regular data. We need all that context to understand why the most common model, ChatGPT, often hallucinates. How AI Hallucinations Happen ChatGPT runs on the Transformer architecture, first introduced in the 2017 paper, “Attention Is All You Need.”  This is the very mechanism that revolutionized text processing. Transformers operate on the self-attention mechanism, which allows them to consider the global context rather than just the nearest words like older recurrent neural networks (LSTM and GRU) do. The model belongs to the GPT (Generative Pre-Trained Transformer) series, which means: Pre-trained: It was initially trained on enormous amounts of text (books, articles, websites, and code).
Generative: Its task is to generate text, not just classify or extract facts. ChatGPT’s answers result from a stochastic process rather than a rigid rule. It doesn’t memorize or reproduce texts but generates responses using a probabilistic model. Word Prediction as a Probabilistic Process When ChatGPT responds, it doesn’t choose the single correct word but computes a probability distribution. P(wi|w1, w2, ..., wi-1), where: “wi” — the next word in the sentence w1, w2, ..., wi-1 — the previous words P(wi|w1, ..., wi-1) — the probability that “wi” will be the next word For example, if you ask, “What day is it today?” ChatGPT might have different probabilities: “Monday” — P=0.7
“Wednesday” — P=0.2
“42” — P=0.0001 It will mostly often choose the word with the highest probability, but due to generation temperature (a parameter that controls randomness), it might sometimes choose a less likely option based on context. Context Influence and Information Forgetting ChatGPT works with a limited context window, meaning it only "remembers" the last NN tokens. For GPT-4, the context window is about 128k tokens (around 300 pages of text). If important information is outside this context, it may: Forget details (context clipping effect)
Make-up information (stochastic process) Yet, ChatGPT can often correct its answer after you ask if it's sure. However, ChatGPT can often correct its answer if you ask whether it’s sure. AI Sometimes Corrects Itself, But Why? When you ask ChatGPT, “Are you sure?” it reanalyzes its answer using a new context where doubt is present. This results in: Recalculating answer probabilities.
Choosing a more plausible option if one exists. This process can be explained by Bayesian probability. P(A|B) = P(B|A)P(A) / P(B), where: P(A|B) — the probability that answer A is correct, considering your follow-up question B.


P(B|A) — the probability that you would have asked if ChatGPT was initially right.


P(A) — the initial probability of ChatGPT's answer.


P(B) — the overall probability that you would ask. Too much information for you? Brain overheating? Imagine that AIs also get overwhelmed by large amounts of information. Errors Due to Overfitting and Noisy Data Massive amounts of text data flow into ChatGPT’s training, including noise or contradictory information, such as: Some sources say the Earth is round, while others claim it’s flat. AI can’t always determine which information is true when it appears with varying probabilities. These are examples of model hallucinations, which occur because ChatGPT’s weights are trained on probabilistic word associations rather than strict logic. The Bottom Line Here is what we can learn from this. ChatGPT hallucinates since it: Predicts probabilistically, not deterministically.


Has a limited memory (context window).


Recalculates probabilities when questioned.


Has training data that includes noise and contradictions. It’s that straightforward. Hope you didn’t get tired. If you did, that’s a good sign because it means you’re thinking critically, which is exactly what we should do when working with AI. For a long time, IT specialists worked without a care in the world. They developed, built, and deployed software smoothly. Then the era of isolation hit, and suddenly, they got bored (of course, this is a playful take on the actual events). IT folks wanted to create something that could handle their work while they stayed home: answer routine questions, generate cool avatars, and analyze vast amounts of data in minutes. They dreamed of traveling to a fantastic place, and so, you guessed it, they revolutionized AI. AI is now functioning, providing answers, and improving lives. As skilled an assistant as it is, AI is truly effective only when used in the right context. We’re witnessing rapid progress in AI applications, from image and video generation to stock market forecasting and cryptocurrency analysis. Yet, AI may offer information we don’t ask for or provide blatantly false answers. Its behavior is very much like that of household cats — you know, the kind that sits quietly and then suddenly pounces on you? AI may offer information we don’t ask for Our cats, as well as AI, enjoy being unpredictable: You give them the same food (or data) — sometimes they eat, sometimes they ignore it. You train them to respond, but they only occasionally react when you call them. The bigger and wilder the cat or the larger the AI model, the harder it is to predict its behavior. In the morning, cats might be calm; by evening, they turn hyperactive (just like dynamic data). Cats might be friendly (deterministic) but can scratch you without warning (stochastic). You give them the same food (or data) — sometimes they eat, sometimes they ignore it. You train them to respond, but they only occasionally react when you call them. The bigger and wilder the cat or the larger the AI model, the harder it is to predict its behavior. In the morning, cats might be calm; by evening, they turn hyperactive (just like dynamic data). Cats might be friendly (deterministic) but can scratch you without warning (stochastic). You might wonder what determinism and stochasticity mean — let’s find out. Determinism and Stochasticity Determinism and Stochasticity A deterministic system always produces the same result given the same input — think idempotency if you're a DevOps engineer . A real-world example would be your cat that eats the same amount of food you put in its bowl every time — this is determinism . But when the cat sniffs and only eats half, it’s no longer deterministic. idempotency DevOps engineer determinism A stochastic process includes an element of randomness: with the same input, the result can vary. For example, a machine learning model often uses stochastic algorithms, like Stochastic Gradient Descent (SGD) , which trains the model by picking random chunks of data rather than the entire dataset. stochastic Stochastic Gradient Descent (SGD) These definitions don’t fully explain why our AIs sometimes hallucinate and behave chaotically. There are other contributing factors, including the following: Determinism Stochasticity Rounding errors and floating-point arithmetic Multithreading and parallel computations Continuously updating data Chaos and the “butterfly effect” Determinism Stochasticity Rounding errors and floating-point arithmetic Multithreading and parallel computations Continuously updating data Chaos and the “butterfly effect” If we look a little closer, we'll see other mechanisms that influence the unpredictable behavior of AI models. A Glimpse of Neural Networks A Glimpse of Neural Networks You probably know that the AIs everyone uses rely on various neural network algorithms. Here are some types of neural networks: AIs everyone uses Fully Connected Neural Networks (FCNN): A classic architecture where each neuron connects to every neuron in the next layer. Fully Connected Neural Networks (FCNN): A classic architecture where each neuron connects to every neuron in the next layer. Fully Connected Neural Networks (FCNN): Convolutional Neural Networks (CNNs): These networks use convolutions or filters that highlight image features like edges, textures, and shapes. Convolutional Neural Networks (CNNs): These networks use convolutions or filters that highlight image features like edges, textures, and shapes. Convolutional Neural Networks (CNNs): Recurrent Neural Networks (RNNs): These networks have feedback loops that allow them to remember previous steps (namely, they remember sequences). Recurrent Neural Networks (RNNs ): These networks have feedback loops that allow them to remember previous steps (namely, they remember sequences). Recurrent Neural Networks (RNNs Long Short-Term Memory (LSTM): An enhanced version of RNNs with mechanisms for selectively forgetting and remembering important data. Long Short-Term Memory (LSTM): An enhanced version of RNNs with mechanisms for selectively forgetting and remembering important data. Long Short-Term Memory (LSTM): Transformers: The most powerful class for text processing. They use multi-head attention, allowing them to consider the entire context simultaneously. Transformers : The most powerful class for text processing. They use multi-head attention, allowing them to consider the entire context simultaneously. Transformers Generative Adversarial Networks (GANs): They consist of two networks, one of which generates data and the other evaluates its quality. Their competition leads to better results. Generative Adversarial Networks (GANs): They consist of two networks, one of which generates data and the other evaluates its quality. Their competition leads to better results. Generative Adversarial Networks (GANs): Autoencoders: Networks designed to compress (encode) information and then reconstruct (decode) it. Autoencoders : Networks designed to compress (encode) information and then reconstruct (decode) it. Autoencoders Graph Neural Networks (GNNs): They work with graphs (nodes and edges) rather than regular data. Graph Neural Networks (GNNs): They work with graphs (nodes and edges) rather than regular data. Graph Neural Networks (GNNs): We need all that context to understand why the most common model, ChatGPT, often hallucinates. How AI Hallucinations Happen How AI Hallucinations Happen ChatGPT runs on the Transformer architecture, first introduced in the 2017 paper, “Attention Is All You Need.” This is the very mechanism that revolutionized text processing. Transformers operate on the self-attention mechanism, which allows them to consider the global context rather than just the nearest words like older recurrent neural networks (LSTM and GRU) do. The model belongs to the GPT (Generative Pre-Trained Transformer) series, which means: Transformer the 2017 paper, “Attention Is All You Need.” Pre-trained: It was initially trained on enormous amounts of text (books, articles, websites, and code). Generative: Its task is to generate text, not just classify or extract facts. Pre-trained: It was initially trained on enormous amounts of text (books, articles, websites, and code). Pre-trained: Generative: Its task is to generate text, not just classify or extract facts. Generative: ChatGPT’s answers result from a stochastic process rather than a rigid rule. It doesn’t memorize or reproduce texts but generates responses using a probabilistic model. Word Prediction as a Probabilistic Process Word Prediction as a Probabilistic Process When ChatGPT responds, it doesn’t choose the single correct word but computes a probability distribution. P(wi|w1, w2, ..., wi-1), where: P(wi|w1, w2, ..., wi-1), “wi” — the next word in the sentence “wi” — the next word in the sentence w1, w2, ..., wi-1 — the previous words P(wi|w1, ..., wi-1) — the probability that “wi” will be the next word P(wi|w1, ..., wi-1) — the probability that “wi” will be the next word For example, if you ask, “What day is it today?” ChatGPT might have different probabilities: “Monday” — P=0.7 “Wednesday” — P=0.2 “42” — P=0.0001 “Monday” — P=0.7 “Wednesday” — P=0.2 “42” — P=0.0001 It will mostly often choose the word with the highest probability, but due to generation temperature (a parameter that controls randomness), it might sometimes choose a less likely option based on context. Context Influence and Information Forgetting Context Influence and Information Forgetting ChatGPT works with a limited context window, meaning it only "remembers" the last NN tokens. For GPT-4, the context window is about 128k tokens (around 300 pages of text). If important information is outside this context, it may: Forget details (context clipping effect) Make-up information (stochastic process) Forget details (context clipping effect) Make-up information (stochastic process) Yet, ChatGPT can often correct its answer after you ask if it's sure. However, ChatGPT can often correct its answer if you ask whether it’s sure. AI Sometimes Corrects Itself, But Why? AI Sometimes Corrects Itself, But Why? When you ask ChatGPT, “Are you sure?” it reanalyzes its answer using a new context where doubt is present. This results in: Recalculating answer probabilities. Choosing a more plausible option if one exists. Recalculating answer probabilities. Choosing a more plausible option if one exists. This process can be explained by Bayesian probability. P(A|B) = P(B|A)P(A) / P(B), where: P(A|B) = P(B|A)P(A) / P(B), P(A|B) — the probability that answer A is correct, considering your follow-up question B. P(B|A) — the probability that you would have asked if ChatGPT was initially right. P(A) — the initial probability of ChatGPT's answer. P(B) — the overall probability that you would ask. P(A|B) — the probability that answer A is correct, considering your follow-up question B. P(A|B) — the probability that answer A is correct, considering your follow-up question B. P(B|A) — the probability that you would have asked if ChatGPT was initially right. P(B|A) — the probability that you would have asked if ChatGPT was initially right. P(A) — the initial probability of ChatGPT's answer. P(A) — the initial probability of ChatGPT's answer. P(B) — the overall probability that you would ask. P(B) — the overall probability that you would ask. Too much information for you? Brain overheating? Imagine that AIs also get overwhelmed by large amounts of information. Errors Due to Overfitting and Noisy Data Errors Due to Overfitting and Noisy Data Massive amounts of text data flow into ChatGPT’s training, including noise or contradictory information, such as: Some sources say the Earth is round, while others claim it’s flat. Some sources say the Earth is round, while others claim it’s flat. AI can’t always determine which information is true when it appears with varying probabilities. AI can’t always determine which information is true when it appears with varying probabilities. These are examples of model hallucinations, which occur because ChatGPT’s weights are trained on probabilistic word associations rather than strict logic. The Bottom Line The Bottom Line Here is what we can learn from this. ChatGPT hallucinates since it: Predicts probabilistically, not deterministically. Has a limited memory (context window). Recalculates probabilities when questioned. Has training data that includes noise and contradictions. Predicts probabilistically, not deterministically. Predicts probabilistically, not deterministically. Has a limited memory (context window). Has a limited memory (context window). Recalculates probabilities when questioned. Recalculates probabilities when questioned. Has training data that includes noise and contradictions. Has training data that includes noise and contradictions. It’s that straightforward. Hope you didn’t get tired. If you did, that’s a good sign because it means you’re thinking critically, which is exactly what we should do when working with AI.

AI’s Non-Determinism, Hallucinations, And... Cats?

Недэтэрмінізм штучнага інтэлекту, галюцынацыі і... кошкі?

Недетерминизмът на AI, халюцинациите и... котките?

AI এর অ-নিয়ন্ত্রণবাদ, হ্যালুসিনেশন, এবং... বিড়াল?

El no determinismo de la IA, las alucinaciones y... ¿los gatos?

A mesterséges intelligencia nem-determinizmusa, hallucinációi és... macskák?

AI の非決定性、幻覚、そして...猫？

Недетерминизам на АИ, халуцинации и... Мачки?

एआईको अ-निर्धारणवाद, भ्रम, र... बिरालोहरू?

Het non-determinisme, hallucinaties en... katten van AI?

AI’s No Determinismo, Alucinaciones, Y... Misikuna?

AI's Non-Determinism, Hallucinations, na ... Injangwe?

AI's Non-Determinism, Hallucinations, Na ... Paka?

Let me tell you my experience 🔎 What is Causality? Traditional statistical and machine learning methods mostly focus on correlations, but causal models allow researchers to infer mechanisms and predict the effects of interventions. Nevertheless, the discovery of cause-and-effect relationships has been always attractive. Indeed, despite being used in many types of data, this is particularly relevant for time series data, for example, the stock market or clinical data as the cerebral activity. Non-time series cases can be based on causal networks of discrete events where one event causes another or more: Flu  →  Fever ↘  →  Cough Rain → Slippery Road → People falling A user Likes Sci-Fi → Recommend him to buy or watch "Interstellar" Those might appear trivial chains, but they can be expanded to many components (also called larger networks) providing insightful inputs that can only be reached computationally given the large amount of data. With the rise of machine learning and deep learning, the need for robust causal inference tools is growing even more compared to the past. Most models from the nearly century-old autoregressors or the more recent zero-shot large foundation models (like TimesFM, CHRONOS, LLMTime)—focus on forecasting values. But they often leave one burning question unanswered: "Does one time series actually influence another?" For example, consider the surge in demand for Nvidia cards. A time series prediction could be able to predict rising stock prices for Nvidia, but what about the ripple effects on semiconductor supplier shares like TSMC? And what if the changes there will be even more pronounced and therefore more relevant (even if caused by another stock share)? That's where causality-driven AI comes into play. 🛠️From Correlation to Causation Emerging from econometrics, the most common approach is the Granger causality. This was introduced by the British economist Clive Granger more than 5 decades ago, and it has been the go-to technique since it has low computation and is therefore easily scalable. It’s simple: if changes in series X help predict series Y after some time, then X is said to “Granger-cause” Y. Here’s a quick Python snippet using the Statsmodels library: import statsmodels.api as sm
from statsmodels.tsa.stattools import grangercausalitytests

# Load sample macroeconomic data
data = sm.datasets.macrodata.load_pandas().data[["realgdp", "realcons"]].pct_change().dropna()
gc_results = grangercausalitytests(data, maxlag=4) The “lag” in the last row is used to indicate how back/far in time we should look for relationships (in brain activity from functional magnetic resonance imaging is generally “1”) While handy, Granger causality is inherently linear. It struggles to capture the non-linear interplay that is so common in complex systems. Obviously, if we are interested in just 2 time series, it is enough. «Complex» means many parts.🚑 🌟 The Stuff Inside EffConPy: EffConPy addresses the aforementioned challenges; it is easily installed by using the command «pip install effconnpy», and it is developed by a team of people led by Dr. Alessandro Crimi from the technical university of Krakow. The library includes multivariate Granger causality, which even if based on Statsmodel, is not out-of-the-box (thank you for putting it together), and more modern approaches such as dynamic Bayesian networks, convergent cross-mapping, etc., which might detect subtle, non-linear-causal-influences-that-traditional models might miss. They are all put together, and therefore, it allows straightforward benchmarks, just changing the method in the last row of this snip, for example: from effconnpy import CausalityAnalyzer
import numpy as np
# Generate sample time series
data = np.random.rand(100, 3)
analyzer = CausalityAnalyzer(data)
results = analyzer.causality_test(method='ccm') If you are wondering why we need something more than Granger causality, let me explain it to you here: Granger causality is explicitly temporal (although some people complain that we have the so-called temporal correlation). It checks whether past values of one variable improve the prediction of another beyond its own past values. However, it assumes linear relationships and does not capture causality in complex dynamical systems where interactions are nonlinear or bidirectional. Dynamic system approaches like Convergent Cross Mapping (CCM) and Transfer Entropy (TE) are focused on those aspects. CCM, based on state-space reconstruction from chaos theory, detects causality by checking if one variable's historical states can reconstruct another’s, implying a shared underlying dynamical system. The math related to this is not just a regression in time. It is based on advanced concepts from the dynamical system of the butterfly effect of the Lorentz attractor: This is nothing esoteric, it is just math particularly useful for feedback systems and ecological interactions. TE, rooted in information theory, quantifies how much knowing one variable reduces uncertainty in another, making it well-suited for detecting nonlinear and asymmetric dependencies. Until now, I have not seen a Python framework putting all this together. 🕵️What I Did NOT Find Inside Effconnpy There has been a growing interest in large foundation models (LFM) for time series prediction, such as TimesFM. Those models are good predictors and they do little work with causality. Therefore, it is understandable that they have been left out by a causal toolbox. However, they are probably the future. Indeed, we need little work to turn them from predictor to causality. Like with Granger, we could check if one series helps predict another one. The other missing thing is reservoir computing causality. A reservoir is not something from your pipe. It is a dynamic, randomly connected network (metaphorically like fluid water), which, in response, to input data, generates intricate patterns. In practice, it is designed by recurrent neural networks. So, after all, it is AI. The patterns created within the «reservoir» help us reveal deeper causal links, such as how fluctuations in one stock market might ripple through related industries or how a stroke disrupts neural connectivity in the brain. This approach was recently introduced in an IEEE Access study titled “End-to-End Stroke Imaging Analysis Using Effective Connectivity and Interpretable Artificial Intelligence” by Wojciech Ciezobka and colleagues, who applied these advanced techniques to healthcare. They used effective connectivity analysis to examine how the activity in one brain region influences another—vital for understanding stroke impacts. It will be great if also this will be included. 📝 Conclusion Causality studies have always been attractive for people interested in time series analysis. By studying the temporal relationships between elements, especially their cause-effect, we allow us to understand better the intricate dynamics that govern our world from the available data, independently if we are talking of economic indicators or the connectivity within the human brain. If you want to know more? As I said, it is open-source and therefore available on the related GitHub repository, or keep the conversation on how causality-driven AI is redefining our understanding of time series data, commenting here.🚀🧠 Let me tell you my experience 🔎 What is Causality? 🔎 What is Causality? Traditional statistical and machine learning methods mostly focus on correlations, but causal models allow researchers to infer mechanisms and predict the effects of interventions. Nevertheless, the discovery of cause-and-effect relationships has been always attractive. Indeed, despite being used in many types of data, this is particularly relevant for time series data, for example, the stock market or clinical data as the cerebral activity. Non-time series cases can be based on causal networks of discrete events where one event causes another or more: Flu  →  Fever Flu  →  Fever ↘  →  Cough ↘  →  Cough Rain → Slippery Road → People falling Rain → Slippery Road → People falling A user Likes Sci-Fi → Recommend him to buy or watch "Interstellar" A user Likes Sci-Fi → Recommend him to buy or watch "Interstellar" Those might appear trivial chains, but they can be expanded to many components (also called larger networks) providing insightful inputs that can only be reached computationally given the large amount of data. With the rise of machine learning and deep learning, the need for robust causal inference tools is growing even more compared to the past. Most models from the nearly century-old autoregressors or the more recent zero-shot large foundation models (like TimesFM , CHRONOS , LLMTime )—focus on forecasting values. But they often leave one burning question unanswered: "Does one time series actually influence another?" TimesFM CHRONOS LLMTime "Does one time series actually influence another?" For example, consider the surge in demand for Nvidia cards. A time series prediction could be able to predict rising stock prices for Nvidia, but what about the ripple effects on semiconductor supplier shares like TSMC? And what if the changes there will be even more pronounced and therefore more relevant (even if caused by another stock share)? That's where causality-driven AI comes into play. 🛠️From Correlation to Causation Emerging from econometrics, the most common approach is the Granger causality . This was introduced by the British economist Clive Granger more than 5 decades ago, and it has been the go-to technique since it has low computation and is therefore easily scalable. It’s simple: if changes in series X help predict series Y after some time, then X is said to “Granger-cause” Y . Granger causality if changes in series X help predict series Y after some time, then X is said to “Granger-cause” Y Here’s a quick Python snippet using the Statsmodels library: import statsmodels.api as sm
from statsmodels.tsa.stattools import grangercausalitytests

# Load sample macroeconomic data
data = sm.datasets.macrodata.load_pandas().data[["realgdp", "realcons"]].pct_change().dropna()
gc_results = grangercausalitytests(data, maxlag=4) import statsmodels.api as sm
from statsmodels.tsa.stattools import grangercausalitytests

# Load sample macroeconomic data
data = sm.datasets.macrodata.load_pandas().data[["realgdp", "realcons"]].pct_change().dropna()
gc_results = grangercausalitytests(data, maxlag=4) The “lag” in the last row is used to indicate how back/far in time we should look for relationships (in brain activity from functional magnetic resonance imaging is generally “1”) While handy, Granger causality is inherently linear. It struggles to capture the non-linear interplay that is so common in complex systems. Obviously, if we are interested in just 2 time series, it is enough. «Complex» means many parts.🚑 🌟 The Stuff Inside EffConPy: EffConPy addresses the aforementioned challenges; it is easily installed by using the command «pip install effconnpy», and it is developed by a team of people led by Dr. Alessandro Crimi from the technical university of Krakow. The library includes multivariate Granger causality, which even if based on Statsmodel, is not out-of-the-box (thank you for putting it together), and more modern approaches such as dynamic Bayesian networks, convergent cross-mapping, etc., which might detect subtle, non-linear-causal-influences-that-traditional models might miss. They are all put together, and therefore, it allows straightforward benchmarks, just changing the method in the last row of this snip, for example: from effconnpy import CausalityAnalyzer
import numpy as np
# Generate sample time series
data = np.random.rand(100, 3)
analyzer = CausalityAnalyzer(data)
results = analyzer.causality_test(method='ccm') from effconnpy import CausalityAnalyzer
import numpy as np
# Generate sample time series
data = np.random.rand(100, 3)
analyzer = CausalityAnalyzer(data)
results = analyzer.causality_test(method='ccm') If you are wondering why we need something more than Granger causality, let me explain it to you here: Granger causality is explicitly temporal (although some people complain that we have the so-called temporal correlation). It checks whether past values of one variable improve the prediction of another beyond its own past values. However, it assumes linear relationships and does not capture causality in complex dynamical systems where interactions are nonlinear or bidirectional. Dynamic system approaches like Convergent Cross Mapping (CCM) and Transfer Entropy (TE) are focused on those aspects. CCM, based on state-space reconstruction from chaos theory, detects causality by checking if one variable's historical states can reconstruct another’s, implying a shared underlying dynamical system. The math related to this is not just a regression in time. It is based on advanced concepts from the dynamical system of the butterfly effect of the Lorentz attractor: Convergent Cross Mapping (CCM) Transfer Entropy (TE) This is nothing esoteric, it is just math particularly useful for feedback systems and ecological interactions. TE, rooted in information theory, quantifies how much knowing one variable reduces uncertainty in another, making it well-suited for detecting nonlinear and asymmetric dependencies. Until now, I have not seen a Python framework putting all this together. 🕵️ What I Did NOT Find Inside Effconnpy 🕵️ What I Did NOT Find Inside Effconnpy NOT There has been a growing interest in large foundation models (LFM) for time series prediction, such as TimesFM . Those models are good predictors and they do little work with causality. Therefore, it is understandable that they have been left out by a causal toolbox. However, they are probably the future. Indeed, we need little work to turn them from predictor to causality. Like with Granger, we could check if one series helps predict another one. TimesFM The other missing thing is reservoir computing causality. A reservoir is not something from your pipe. It is a dynamic, randomly connected network (metaphorically like fluid water), which, in response, to input data, generates intricate patterns. In practice, it is designed by recurrent neural networks. So, after all, it is AI. reservoir computing causality. The patterns created within the «reservoir» help us reveal deeper causal links, such as how fluctuations in one stock market might ripple through related industries or how a stroke disrupts neural connectivity in the brain. This approach was recently introduced in an IEEE Access study titled “End-to-End Stroke Imaging Analysis Using Effective Connectivity and Interpretable Artificial Intelligence” by Wojciech Ciezobka and colleagues , who applied these advanced techniques to healthcare. They used effective connectivity analysis to examine how the activity in one brain region influences another—vital for understanding stroke impacts. It will be great if also this will be included. IEEE Access study titled “End-to-End Stroke Imaging Analysis Using Effective Connectivity and Interpretable Artificial Intelligence” by Wojciech Ciezobka and colleagues “End-to-End Stroke Imaging Analysis Using Effective Connectivity and Interpretable Artificial Intelligence” 📝 Conclusion Causality studies have always been attractive for people interested in time series analysis. By studying the temporal relationships between elements, especially their cause-effect, we allow us to understand better the intricate dynamics that govern our world from the available data, independently if we are talking of economic indicators or the connectivity within the human brain. If you want to know more? As I said, it is open-source and therefore available on the related GitHub repository , or keep the conversation on how causality-driven AI is redefining our understanding of time series data, commenting here.🚀🧠 the related GitHub repository

EffConPy: Open Source Causality Discovery in Python

Nature + Biomimicry + Bitcoin: Mutually Beneficial Partnership

Naturaleza + Biomimetismo + Bitcoin: una alianza mutuamente beneficiosa

Luonto + Biomimikri + Bitcoin: molemminpuolisesti hyödyllinen kumppanuus

प्रकृति + बायोमिमिक्री + बिटकॉइन: पारस्परिक रूप से लाभकारी साझेदारी

Nature + Biomimicry + Bitcoin. փոխշահավետ գործընկերություն

自然 + バイオミミクリー + ビットコイン: 相互に利益のあるパートナーシップ

Nature + Biomimicry + Bitcoin: ორმხრივად მომგებიანი პარტნიორობა

ធម្មជាតិ + Biomimicry + Bitcoin៖ ភាពជាដៃគូដែលមានអត្ថប្រយោជន៍ទៅវិញទៅមក

Natiora + Biomimicry + Bitcoin: Fiaraha-miombon'antoka

Masikirwo + Biomimicry + Bitcoin: Kudyidzana Kunobatsira Kudyidzana

Природа + биомимикрија + биткоин: обострано корисно партнерство

இயற்கை + பயோமிமிக்ரி + பிட்காயின்: பரஸ்பர நன்மை பயக்கும் கூட்டாண்மை

Tabiat + Biomimicry + Bitcoin: O'zaro manfaatli hamkorlik

Sales Pitch Shake-Up: Part Two—The Pitch Hack So Obvious, You’ll Kick Yourself

The above image is a beautiful image of a butterfly meshed in vibrant colors with musical notes all across it. With so much going on in it, what one person sees in it (at least at first) can be very different from another person. A little while ago, I was working on a neural style transfer project. This is one of the images that I used as the style image. However, the resultant image didn’t display as much difference from the original base image as I’d thought. Even so, I gaily showed it to a couple of friends of mine. I asked what they thought of it, and they said that the base image and the butterfly image both had a significant chunk of white in them which was why the color transfer wasn’t obvious. I honestly stared at the image for a couple of minutes before I realized that what they had referred to as the ‘butterfly image’ was in fact my style image. See, the thing that had initially attracted me to the picture was the color explosion in combination with the musical notes and the writings. It was all so electric. But this wasn’t what my friends saw at first glance, and like I said, I didn’t see the butterfly until they pointed it to me even though, from a logical perspective it really should have been. The point of this entire rambling is that people have different perspectives, and when multiple people are working on a project, in order to make it a success, each individual’s perspective and way of thinking should be clear to the rest of the team. This is the definition of coordination in a functional system. Achieving this level of coordination is hard enough face-to-face, but it doubles down on the internet. That, to me, seems to be one of the issues plaguing the decentralization community. What Is a Blockchain Network? I will give a brief overview of what decentralization and blockchain networks are so that everyone can understand my point more clearly. The blockchain is like a ledger that everyone can read - at least if it’s public. If it’s private or semi-public, then you have to be one of the authorized people to be able to view and change anything in the ledger. But even for public blockchains, a transaction performed on it needs to be verified. The more times it is verified, the more its authenticity is assured. The trouble is these verifications require solving complex mathematical problems which, in turn, require vast amounts of computational resources. Moreover, people working on these are individuals who do not, in real life, know each other. They are only connected through their mutual interest in decentralization. Therefore, in order to ascertain cohesive coordination among these individuals to assure the formation of a well-coordinated decentralized system, protocols are established by the network administrators to clarify the vision and mission of the network. Now understandably, most of these networks’ missions and protocols revolve around data privacy and security because from what I’ve been able to understand--courtesy of Google’s Gemini--is that it is completely possible for a group of hackers or cybercriminals to create a private blockchain network of their own where they freely store and share all kinds of hacked data. So clearly, the blockchain still needs some work to rid it of criminal activity. Now, I understand that my example may be a little extreme and that law-enforcement agencies probably have similar sophisticated technologies to counteract these crimes, and things like digital fingerprints exist, as judiciously explained by Gemini. Nevertheless, these are problems that the blockchain community is rightfully interested in resolving. But that leads to another problem: tunnel vision. The Big Picture The blockchain community is a very exclusive community, composed mostly of young, enthusiastic, tech-savvy individuals. However, the idea of data privacy and security is for everyone, not just the techies. So, it is important that this technology is used and understood as much by regular individuals as it is by its developers. Imagine this scenario. A bank manager or administrator of a small, local bank decides that he wants to convert his banking system to a decentralized one to better the security for the bank customers. He hires a young developer to help and guide him through this change. The developer talks to him about authentication codes, verification, mining data, and other resources needed to form an effective blockchain network. The manager--even if he is relatively young--will probably understand that the system access should be limited to a few trusted individuals, and maybe a little bit more. However, unless he himself is a computer scientist--unlikely in the case of a bank manager--it’s doubtful that he will understand the whole concept of decentralization. But the tipping point would be the computational costs required. Remember, the bank is a small, local bank with a limited budget and resources. When you add this to the fact that the manager may or may not have completely understood how the system works, the odds that the bank will end up installing an effective decentralized system are slim. The Gaps One of the main issues stopping decentralization from becoming a global phenomenon is the lack of suitable hardware, closely followed by inadequate education for the masses. My neural style transfer project was actually not the project I’d originally meant to do. Instead, I had intended to create a Generative AI project using transfer learning to make a GAN algorithm create some new images. My dataset--which was composed of the works of lesser-known artists--wasn’t a very big one, so I knew that it was possible that the end result may not be as novel as it would have been with a bigger dataset. But what I hadn’t realized was that even with that small dataset the time, space, and computation required would be so enormous that, for my purposes, it would be better if I switched to a slightly different type of algorithm. That’s kind of my point. I use a regular laptop for most of my work, and for the most part, even for other A.I. projects, it’s usually enough. But it seems that for projects that require any kind of ‘generation’--even to a small extent--this can be a very big problem. In my case, it wasn’t that big of a deal since it was more of a personal project than anything else, but for the task of providing privacy and security to the clients of a bank, this may be a bigger issue because while their computational resources may be better than my tiny, lonely laptop, it’s still unlikely to have the computational power even remotely resembling that which is necessary for maintaining blockchain networks. Furthermore, the blockchain community really needs to understand that most people in positions responsible for the privacy and security of organizations are at the very least in their forties. Even if they are younger, they’re probably not very aware of how decentralization systems work which is understandable given its complexity. One of the reasons that social media platforms are so universally popular across all demographics is that aside from knowing what a few essential buttons do, people don’t need to know very much about it. This is especially so in cases of individuals over the age of 60, and while your average bank manager may not be 60, he/she isn’t going to be 20 or even 30 either. Conclusion There is a lot of money being spent on making decentralized systems more reliable and eco-friendly, but maybe what is necessary is some of that money to also be spent on creating hardware that is not so costly and on educating individuals, i.e., marketing according to their demographic and positions in life. I mean, I can use my social media account from any phone and the phone needn’t be a good quality one, or even a new one. As long as the wifi and my app work, I don’t need anything else. All in all, ensuring online coordination requires in and of itself an extensive system. If you’re only looking at the shape of the butterfly then you’re missing the colors and beauty associated with it, and if you only look at the colors, you might end up missing the main thing. So, maybe what the blockchain and decentralization community needs right now for optimal coordination and cooperation are not just networks that focus on increasing privacy and security but instead a system of network of networks, where one network may be responsible for ensuring privacy and security, the other for hardware, and still other which works on marketing and educating plans for the masses all connected towards one unified goal. In this way, privacy and security will no longer be just a catchphrase for the tech-savvy but an asset utilized by everyone. The above image is a beautiful image of a butterfly meshed in vibrant colors with musical notes all across it. With so much going on in it, what one person sees in it (at least at first) can be very different from another person. A little while ago, I was working on a neural style transfer project. This is one of the images that I used as the style image. However, the resultant image didn’t display as much difference from the original base image as I’d thought. Even so, I gaily showed it to a couple of friends of mine. I asked what they thought of it, and they said that the base image and the butterfly image both had a significant chunk of white in them which was why the color transfer wasn’t obvious. I honestly stared at the image for a couple of minutes before I realized that what they had referred to as the ‘butterfly image’ was in fact my style image. See, the thing that had initially attracted me to the picture was the color explosion in combination with the musical notes and the writings. It was all so electric. But this wasn’t what my friends saw at first glance, and like I said, I didn’t see the butterfly until they pointed it to me even though, from a logical perspective it really should have been. The point of this entire rambling is that people have different perspectives, and when multiple people are working on a project, in order to make it a success, each individual’s perspective and way of thinking should be clear to the rest of the team. This is the definition of coordination in a functional system. Achieving this level of coordination is hard enough face-to-face, but it doubles down on the internet. That, to me, seems to be one of the issues plaguing the decentralization community. What Is a Blockchain Network? I will give a brief overview of what decentralization and blockchain networks are so that everyone can understand my point more clearly. The blockchain is like a ledger that everyone can read - at least if it’s public. If it’s private or semi-public, then you have to be one of the authorized people to be able to view and change anything in the ledger. But even for public blockchains, a transaction performed on it needs to be verified. The more times it is verified, the more its authenticity is assured. The trouble is these verifications require solving complex mathematical problems which, in turn, require vast amounts of computational resources. Moreover, people working on these are individuals who do not, in real life, know each other. They are only connected through their mutual interest in decentralization. Therefore, in order to ascertain cohesive coordination among these individuals to assure the formation of a well-coordinated decentralized system, protocols are established by the network administrators to clarify the vision and mission of the network. Now understandably, most of these networks’ missions and protocols revolve around data privacy and security because from what I’ve been able to understand--courtesy of Google’s Gemini--is that it is completely possible for a group of hackers or cybercriminals to create a private blockchain network of their own where they freely store and share all kinds of hacked data. So clearly, the blockchain still needs some work to rid it of criminal activity. Now, I understand that my example may be a little extreme and that law-enforcement agencies probably have similar sophisticated technologies to counteract these crimes, and things like digital fingerprints exist, as judiciously explained by Gemini. Nevertheless, these are problems that the blockchain community is rightfully interested in resolving. But that leads to another problem: tunnel vision. The Big Picture The blockchain community is a very exclusive community, composed mostly of young, enthusiastic, tech-savvy individuals. However, the idea of data privacy and security is for everyone, not just the techies. So, it is important that this technology is used and understood as much by regular individuals as it is by its developers. Imagine this scenario. A bank manager or administrator of a small, local bank decides that he wants to convert his banking system to a decentralized one to better the security for the bank customers. He hires a young developer to help and guide him through this change. The developer talks to him about authentication codes, verification, mining data, and other resources needed to form an effective blockchain network. The manager--even if he is relatively young--will probably understand that the system access should be limited to a few trusted individuals, and maybe a little bit more. However, unless he himself is a computer scientist--unlikely in the case of a bank manager--it’s doubtful that he will understand the whole concept of decentralization. But the tipping point would be the computational costs required. Remember, the bank is a small, local bank with a limited budget and resources. When you add this to the fact that the manager may or may not have completely understood how the system works, the odds that the bank will end up installing an effective decentralized system are slim. The Gaps One of the main issues stopping decentralization from becoming a global phenomenon is the lack of suitable hardware, closely followed by inadequate education for the masses. My neural style transfer project was actually not the project I’d originally meant to do. Instead, I had intended to create a Generative AI project using transfer learning to make a GAN algorithm create some new images. My dataset--which was composed of the works of lesser-known artists--wasn’t a very big one, so I knew that it was possible that the end result may not be as novel as it would have been with a bigger dataset. But what I hadn’t realized was that even with that small dataset the time, space, and computation required would be so enormous that, for my purposes, it would be better if I switched to a slightly different type of algorithm. That’s kind of my point. I use a regular laptop for most of my work, and for the most part, even for other A.I. projects, it’s usually enough. But it seems that for projects that require any kind of ‘generation’--even to a small extent--this can be a very big problem. In my case, it wasn’t that big of a deal since it was more of a personal project than anything else, but for the task of providing privacy and security to the clients of a bank, this may be a bigger issue because while their computational resources may be better than my tiny, lonely laptop, it’s still unlikely to have the computational power even remotely resembling that which is necessary for maintaining blockchain networks. Furthermore, the blockchain community really needs to understand that most people in positions responsible for the privacy and security of organizations are at the very least in their forties. Even if they are younger, they’re probably not very aware of how decentralization systems work which is understandable given its complexity. One of the reasons that social media platforms are so universally popular across all demographics is that aside from knowing what a few essential buttons do, people don’t need to know very much about it. This is especially so in cases of individuals over the age of 60, and while your average bank manager may not be 60, he/she isn’t going to be 20 or even 30 either. Conclusion Conclusion There is a lot of money being spent on making decentralized systems more reliable and eco-friendly, but maybe what is necessary is some of that money to also be spent on creating hardware that is not so costly and on educating individuals, i.e., marketing according to their demographic and positions in life. I mean, I can use my social media account from any phone and the phone needn’t be a good quality one, or even a new one. As long as the wifi and my app work, I don’t need anything else. All in all, ensuring online coordination requires in and of itself an extensive system. If you’re only looking at the shape of the butterfly then you’re missing the colors and beauty associated with it, and if you only look at the colors, you might end up missing the main thing. So, maybe what the blockchain and decentralization community needs right now for optimal coordination and cooperation are not just networks that focus on increasing privacy and security but instead a system of network of networks, where one network may be responsible for ensuring privacy and security, the other for hardware, and still other which works on marketing and educating plans for the masses all connected towards one unified goal. In this way, privacy and security will no longer be just a catchphrase for the tech-savvy but an asset utilized by everyone.

What Is Your Perspective?

Stories, News & Mentions about BUTTERFLY

Categories

Trending Topics

Stories, News & Mentions about BUTTERFLY

Butterfly’s Stories on HackerNoon

Categories

Trending Topics