AI World Whitepaper 1.5
Last updated
Last updated
AIworld:
Blockchain-Based AI Aggregation Platform to Make Cutting-Edge Technology More Accessible
Gong Hui 1,2 and Ioannis Korkos, 1,3,4
1 AIworld Foundation Advisor
2 Westminster Business School, University of Westminster
3 Senior lecturer at the University of South London 4 Dept. of Computer Science,
Abstract
Blockchain is a new internet protocol based on distributed ledger technology, with its most significant innovation being its decentralized and tamper-proof characteristics. Meanwhile, artificial intelligence (AI) is one of the fastest-developing technologies in recent years and has been widely applied and developed in many fields. Combining AI and blockchain not only provides more efficient and secure guarantees for the application of blockchain, but also promotes the development and application of AI technology.
This article will explore the integration points, applications, and development prospects of these two technologies from the perspective of blockchain and AI. We will integrate the AI with the highest Turing test completion rate into various AI products, and upgrade its usability and understandability to lower the entry barriers for the public. Additionally, we will reduce the system pressure caused by high-frequency concurrent access to servers by releasing and utilizing users' idle local computing power.
Content
AIworld: 1
NOTICE 4
I. Preface 5
Introduction 5
Purpose 5
II. Overview of Blockchain Technology 6
Principles 6
1.Distributed Ledger 6
2.Chain Structure 7
3.Consensus Mechanism 7
Characteristics 7
1.Decentralization 7
2.Immutability 7
3.Traceability 7
Applications 8
1.Cryptocurrencies 8
2.Smart Contracts 8
3.IoT and other storage/computing/validation fields 8
III. Overview of AI 8
Types and Characteristics of Artificial Intelligence 9
Deep learning and machine learning 10
Artificial Intelligence Applications 11
IV. Integration of Blockchain and AI 13
The Combination of Blockchain and AI Value 13
Use cases of blockchain and AI 14
V. AIworld Advantages and Computing Nodes 15
About Us 15
Deep Reinforcement Learning 15
Capabilities and Features 18
Core team member 20
VI. Technical Challenges of On-Chain AI 23
Collection: Simplified data collection and accessibility. 23
Organization: Creating business-oriented analysis foundations. 23
Analysis and Integration: 24
VII. Services and Advantages of AIworld 24
Intelligent Language Model 24
Graphic Processing Model 25
Video AI Model 26
VIII. Token Distribution 26
Computing Nodes 27
Mining Rewards: 28
Node Price 28
Ecological Construction 29
Token Governance and Consumption 29
Other token usage scenarios and rules 30
NOTICE
NOTHING IN THIS WHITEPAPER CONSTITUTES LEGAL, FINANCIAL, BUSIÂNESS, OR TAX ADVICE AND YOU SHOULD CONSULT YOUR OWN LEGAL, FINANCIAL, TAX OR OTHER PROFESSIONAL ADVISER BEFORE ENGAGING IN ANY ACTIVITY IN CONNECTION HEREWITH. NEITHER AI FOUNÂDATION LTD. ANY OF THE PROJECT TEAM MEMBERS WHO HAVE WORKED ON THE PLATFORM OR PROJECT IN ANY WAY WHATSOEVER NOR ANY THIRD PARTY SERÂVICE PROVIDER SHALL BE LIABLE FOR ANY KIND OF DIRECT OR INDIÂRECT DAMAGE OR LOSS WHATSOEVER WHICH YOU MAY SUFFER IN CONÂNECTION WITH ACCESSING THIS WHITEPAPER, MATERIALS PRODUCED BY THE AIworld, OR ACCESSING THE WEBSITE AT [HTTPS://WWW.AI- CHAIN.ORG/] OR ANY OTHER MATERIALS PUBLISHED BY THE AI.
The AI and the AI team do not and do not purport to make, and hereby disclaims, all representations, warranties or undertaking to any entity or person. All statements contained in this Whitepaper, statements made in press releases or in any place accessible by the public and oral statements that may be made by the AIworld and/or the AIworld team may constitute forward looking statements (including stateÂments regarding intent, belief or current expectations with respect to market conditions, business strategy and plans, financial condition, specific provisions and risk manageÂment practices). You are cautioned not to place undue reliance on these forward-looking statements given that these statements involve known and unknown risks, uncertainties and other factors that may cause the actual future results to be materially different from that described by such forward looking statements. These forward-looking statements are applicable only as of the date of this Whitepaper and the AIworld and the AIworld team expressly disclaims any responsibility (whether express or implied) to release any revisions to these forward-looking statements to reflect events after such date. You understand that the Project and the creation and distribution of the Tokens involve significant risks, including but not limited to, the risk that (i) the technology associated with the AIworld Project may not function as intended; (ii) the AIworld Project may fail to attract interest or adoption, either from key stakeholders or the broader community; (iii) no guarantees that the price per Token determined by the market will be equal to or higher.
Blockchain technology and artificial intelligence (AI) are two disruptive technologies. The decentralization and tamper-proof nature of blockchain technology can support applications such as smart contracts and digital currencies, while the capabilities of AI technology, such as big data processing, automated reasoning, and deep learning, can enable computers to better handle massive amounts of data and provide intelligent analysis and decision-making. The fusion of blockchain technology and AI can not only fully leverage the advantages of both, but also further promote revolutionary changes in the digital economy and social development, further lowering the cost and threshold of ordinary people using blockchain technology and AI, and truly benefiting everyone economically.
The fusion of blockchain technology and AI can promote the development of smart contracts. Smart contracts are automated contracts implemented using blockchain technology, which can automatically execute contract terms, achieve decentralized transactions and transfers, reduce human intervention, and reduce transaction costs. The big data processing and automated reasoning capabilities of AI technology make smart contracts even more intelligent and automated. For example, based on data analyzed by AI technology, smart contracts can automatically decide whether to execute terms or adjust contract terms to better meet market demands.
The fusion of blockchain technology and AI can strengthen digital identity verification and privacy protection. The decentralization and tamper-proof nature of blockchain technology can ensure the authenticity and security of identity verification and transaction records, while AI technology, such as facial recognition, speech recognition, and fingerprint recognition, can verify identity more accurately. At the same time, these AI technologies can also achieve privacy protection, such as using multi-party computing protocols to ensure sensitive data is not leaked, or using encryption technology to achieve data confidentiality.
For this project, the platform developers and underlying AI model builders come from various leading AI teams. This effectively lowers the usage threshold for the general public and utilizes idle computing power to enhance the current level of AI. It promotes a revolutionary change in the digital economy and society. The digital economy and society require more efficient, secure, and intelligent infrastructure and services, and the integration of blockchain technology and AI technology can provide a more complete technical foundation for the digital economy and society. Through the integration of blockchain technology and AI technology, innovation can be achieved in multiple areas, such as digital identity, digital transactions, and digital governance, bringing greater innovation and competitiveness to the development of the digital economy and society.
This aggregation platform has great potential and significant advantages, which can promote the development and innovation of the digital economy and society. Although the application of these technologies is still in the exploratory and practical stages, it can be imagined that in future development, the integration of blockchain technology and AI technology will provide more complete and intelligent digital services, and bring more innovation and value to society.
Blockchain is a distributed, decentralized ledger technology that has received increasing attention and application in recent years. Its basic principle is to link multiple data blocks in sequence to form an immutable and publicly available ledger, thereby ensuring the security and transparency of data.
Blockchain adopts distributed ledger technology, and all nodes have the same data, with each node having the right to confirm transactions on the ledger. If any part of the node has a problem, it will not affect the normal operation of the entire system.
Each block contains the hash value of the previous block, forming a chain-like structure database. The data in each block is verified through hashing, timestamps, and other related mechanisms, ensuring the reliability and integrity of the data.
Transactions in blockchain need to be confirmed and verified by multiple nodes, and this consensus mechanism ensures the correctness and consistency of the data. Commonly used consensus mechanisms include Proof of Work (PoW), Proof of Stake (PoS), and Delegated Proof of Stake (DPoS), etc.
The data storage and transmission of blockchain do not rely on any centralized organization or institution, but are maintained and managed by all nodes in the network together. This decentralization characteristic makes it more secure and reliable without single-point failure or attack problems.
Each block in the blockchain is encrypted and validated by hashing, and each block contains the hash value of the previous block, forming an immutable chain. If someone tries to tamper with the data of a block, the hash values of all subsequent blocks will change, thus preventing data tampering.
All transactions in the blockchain are recorded in the ledger, and these transaction records are public and transparent, and anyone can view them. This traceability characteristic makes blockchain widely applicable in tracing goods, protecting intellectual property rights, and governance and supervision.
Currently, the most widespread application of blockchain technology is in the realm of cryptocurrencies, with Bitcoin being the most famous example. By utilizing blockchain technology, Bitcoin achieves decentralized transactions and asset management, becoming an important form of currency in the digital economy era.
Smart contracts are a new type of contract that combines blockchain technology with artificial intelligence, enabling automated and intelligent contract management on the blockchain. Smart contracts have been widely applied in fields such as supply chain management and asset trading.
Blockchain technology can also be applied in the IoT field. For example, in the agricultural industry, blockchain technology can be used to trace the source and quality of agricultural products, improving food safety; in the industrial sector, blockchain technology can be used to manage equipment maintenance and repair records, improving the ability to troubleshoot equipment failures.
Although various definitions of Artificial Intelligence (AI) have emerged over the past few decades, John McCarthy provided the following definition in his 2004 paper: "It is the science and engineering of making intelligent machines, especially intelligent computer programs. It is related to the similar task of using computers to understand human intelligence, but AI does not have to confine itself to methods that are biologically observable."
However, the birth of AI conversation dates back to Alan Turing's groundbreaking work "Computing Machinery and Intelligence," published in 1950, decades before this definition emerged. In this paper, Turing, often referred to as the "father of computer science," posed the question "Can machines think?" and went on to propose the famous "Turing Test," in which a human examiner tries to distinguish between responses from a computer and those from a human through text. Although the test has undergone significant scrutiny since its publication, it remains an important part of the history of AI and a concept that continues to evolve in philosophy, as it utilizes ideas related to linguistics.
Stuart Russell and Peter Norvig later published "Artificial Intelligence: A Modern Approach", which became one of the main textbooks for AI research. In the book, they discussed four potential goals or definitions of AI, distinguishing AI from computer systems in terms of rationality and thinking and action:
Human-like methods:
Systems that think like humans
Systems that act like humans
Ideal methods:
Systems that think rationally
Systems that act rationally
Alan Turing's definition falls into the category of "systems that act like humans".
In its simplest form, artificial intelligence is a field that combines computer science and powerful datasets to achieve problem-solving. It also includes subfields such as machine learning and deep learning, which are often mentioned together with AI. These disciplines consist of AI algorithms that aim to create expert systems based on input data for prediction or classification.
AI can be divided into two categories - strong AI and weak AI.
Weak AI, also known as narrow AI or Artificial Narrow Intelligence (ANI), is trained AI that focuses on performing specific tasks. Weak AI drives most of the AI around us today. "Narrow in scope" might be a more accurate descriptor for this type of AI, as it is actually not weak and supports some very powerful applications, such as Apple's Siri, Amazon's Alexa, IBM Watson, and self-driving cars.
Strong AI is composed of Artificial General Intelligence (AGI) and Artificial Superintelligence (ASI). AGI is a theoretical form of AI where machines possess intelligence equivalent to humans. It has self-awareness and can solve problems, learn, and plan for the future. ASI, also known as superintelligence, will surpass the intelligence and abilities of the human brain. Although strong AI is still entirely in the theoretical stage and there are no practical examples of it yet, this does not mean that AI researchers are not exploring its development. The best examples of ASI may come from science fiction, such as HAL, Superman, and the rogue computer assistant in the movie "2001: A Space Odyssey".
Deep learning and machine learning are two terms that are often used interchangeably, so it is important to note the subtle differences between them. As mentioned above, both deep learning and machine learning are subfields of artificial intelligence, with deep learning being a subset of machine learning.
Deep learning is actually composed of neural networks. The "depth" in deep learning refers to neural networks that have three or more layers (including input and output) and can be considered as deep learning algorithms. This is typically represented as shown in the following figure:
The difference between deep learning and machine learning lies in how each algorithm learns. Deep learning can automate much of the feature extraction process, eliminating some necessary human intervention, and can work with larger datasets. Deep learning can be thought of as "scalable machine learning," as Lex Fridman pointed out in the same MIT lecture. Conventional machine learning, or "non-deep" machine learning, relies more on human intervention for learning. Human experts determine the hierarchical structure of features to understand differences between data inputs, often requiring more structured data for learning.
"Deep" machine learning can utilize labeled datasets, also known as supervised learning, to determine algorithms, but it doesn't necessarily have to use labeled datasets. It can collect unstructured data in its raw format (such as text or images) and automatically determine a hierarchical structure of features that differentiate different categories of data. Unlike machine learning, it doesn't require human intervention in data processing, allowing us to expand machine learning in more interesting ways.
Currently, there are numerous real-world applications of AI systems. Below are some of the most common examples：
Also known as automatic speech recognition (ASR), computer speech recognition, or speech-to-text, it is capable of using natural language processing (NLP) to convert human speech into written format. Many mobile devices integrate speech recognition into their systems for voice search, such as Siri, or to provide more assistive features about the text.
Online chatbots are gradually replacing human customer service in customer interactions. They answer frequently asked questions (FAQ) on various topics, such as delivery, provide personalized recommendations, cross-sell products, offer sizing advice to users, and have changed the way we view customer interactions on websites and social media. Examples include chatbots with virtual customer service on e-commerce sites, messaging apps such as Slack and Facebook Messenger, and tasks typically performed by virtual assistants and voice assistants.
This AI technology enables computers and systems to extract meaningful information from digital images, videos, and other visual inputs and take action based on those inputs. This ability to provide recommendations distinguishes it from image recognition tasks. Computer vision is supported by convolutional neural networks and is applied in areas such as photo tagging on social media, radiography in healthcare, and self-driving cars in the automotive industry.
AI algorithms use past consumer behavior data to help identify trends in data that can be used to develop more effective cross-selling strategies. This is used to provide relevant additional recommendations to customers during the checkout process at online retailers.
AI-driven high-frequency trading platforms, designed to optimize stock portfolios, can generate thousands to millions of trades per day without human intervention.
Blockchain is a shared, immutable ledger that provides real-time, shared, and transparent encrypted data exchange to multiple parties initiating and completing transactions. Blockchain networks can track orders, payments, accounts, and production, among other things. Due to licensed members having a unified view of facts, they have gained confidence and trust in transactions with other businesses, increasing efficiency and creating new business opportunities.
Artificial intelligence uses computers, data, and sometimes even machines to simulate human problem-solving and decision-making capabilities. It also includes subdomains such as machine learning and deep learning that use data-trained AI algorithms to make predictions or classifications and become smarter over time. The benefits of AI include automating repetitive tasks, improving decision-making, and creating a better customer experience.
The digital records of blockchain provide insights into the underlying framework of AI and its data sources, addressing the challenge of interpretable AI. This helps to enhance trust in data integrity and, in turn, increases trust in the recommendations provided by AI. Using blockchain to store and distribute AI models provides audit tracking, while pairing blockchain and AI can enhance data security.
AI is capable of quickly and comprehensively reading, understanding, and correlating data at an astonishing speed, bringing new levels of intelligence to blockchain-based business networks. By providing access to large amounts of data inside and outside of enterprises, blockchain can help AI achieve scalability, provide more feasible insights, manage data usage and model sharing, and create a trusted and transparent data economy.
AI, automation, and blockchain can bring new value to cross-party business processes by eliminating friction, speeding up processes, and increasing efficiency. For example, AI models embedded in smart contracts executed on blockchain can recommend products to be recalled, execute transactions (such as reordering, payment, or purchasing stocks based on set thresholds and events), resolve disputes, and select the most sustainable shipping methods.
Introducing AI to blockchain in various industries has brought new opportunities to people.
AI can almost help drive every aspect of the healthcare industry, from presenting treatment insights and supporting user needs to identifying insights and patterns from patient data. Through patient data, including electronic health records, shared on blockchain, companies can collaborate to improve the level of care while protecting patient privacy.
The combination of blockchain and AI in the pharmaceutical industry can increase the visibility and traceability of the drug supply chain while significantly improving the success rate of clinical trials. By combining advanced data analytics with a decentralized clinical trial framework, data integrity, transparency, patient tracking, consent management, and automation of trial participation and data collection can be achieved.
By building trust, eliminating friction in multi-party transactions, and speeding up transaction times, blockchain and AI are revolutionizing the financial services industry. Consider the loan application process. Applicants grant permission to access personal records stored on the blockchain. Trust in the data used for evaluating applications and automated processes helps speed up the process and increase customer satisfaction.
AI and blockchain are fundamentally changing various industries' supply chains and creating new business opportunities by digitizing paper-based processes, making data shareable and trustworthy, and adding intelligence and automation to transaction execution. For example, manufacturers can track carbon emission data at the product or component level to improve the accuracy and intelligence of decarbonization efforts.
The AI aggregation platform created by AIWorld is the latest milestone in expanding deep learning, made possible through the efforts of top AI studios such as OpenAI. The platform is a large-scale multimodal model that accepts image and text inputs and outputs text, although its abilities in many real-world scenarios may not match those of humans, it performs at a human level on various professional and academic benchmarks.
We spent sixteen months iterating and adjusting with our adversarial testing programs and lessons learned from ChatGPT/mid to achieve the best results ever in terms of realism, manipulability, and rejection beyond the fence (although far from perfect). Over the past two years, we have rebuilt the entire deep learning stack and worked with Azure to design a supercomputer to support our workloads. We have discovered and fixed some errors and improved our theoretical foundations.
Currently, our model is more stable than ever before, becoming the first large-scale model for which we can accurately predict its training performance in advance. As we continue to focus on reliable scaling, our goal is to refine our methods to help us predict and prepare for future capabilities earlier - which we believe is critical to safety.
Deep learning is a type of machine learning that trains artificial neural networks to transform a set of inputs into a specific set of outputs. Deep learning often takes the form of supervised learning, using labeled datasets for training. The approach of deep learning can directly handle high-dimensional and complex raw input data, and does not require manual feature engineering to extract features from input data, as compared to previous methods. Therefore, deep learning has brought breakthrough progress in fields such as computer vision and natural language processing.
On the other hand, reinforcement learning involves making intelligent agents interact with environments and trying to learn from the errors to make better decisions. Such problems are often represented mathematically using Markov decision processes, emphasizing how to act based on the environment to achieve maximum expected benefits. Unlike supervised learning, reinforcement learning does not require labeled input-output pairs, and does not need to precisely correct non-optimal solutions.
AIworld's reinforcement deep learning uses deep learning techniques to solve decision-making problems in reinforcement learning, training artificial neural networks to develop specialized algorithms for different training scenarios (such as natural language modeling and humanoid visual processing).
Compared to OpenAI's CLIP (Contrastive Language-Image Pre-Training), AIworld's model is an advanced deep learning model that can understand text and images. It is a multimodal model that has been trained to match text and images, learning to recognize the contents in images and describe the language of images.
The main goal of CLIP is to learn to match images and text through contrastive learning. This is achieved by training the model to predict which image belongs to a given text, and vice versa. During the training process, the model learns to encode images and text into a unified vector space, which enables it to understand the relationship between them in both language and vision. In this way, CLIP can recognize elements such as objects, scenes, and actions in images, while also understanding text related to the images such as labels, descriptions, and titles. CLIP has been shown to perform excellently on visual and language tasks. It is based on an adversarial learning image classification model, which can simultaneously understand natural language descriptions and image content, and establish connections between the two.
The adversarial learning approach used by CLIP is similar to that of GANs (Generative Adversarial Networks). It consists of an image encoder and a text encoder, which are trained through adversarial learning. During adversarial training, the image encoder attempts to minimize the distance between images and text, while the text encoder tries to maximize the distance between them. This training approach helps the model learn better representations of images and text, enabling it to perform excellently on a variety of visual and language tasks.
The CLIP model performs exceptionally well on various image classification tasks, including general image classification, fine-grained classification, retrieval tasks, and support for few-shot learning. Additionally, it can recognize and understand complex concepts within images. Finally, through the development of another generative model based on the CLIP model, it is possible to generate images based on natural language descriptions.
Multimodal with image and text input and text output
Performance: While its performance in real-world scenarios may not match that of humans, it performs at a human level on various professional and academic benchmarks.
Compared to previous models such as ChatGPT, GPT-3.5, and mid, the AI aggregation platform is more reliable, creative, and capable of handling more subtle instructions when the complexity of the task reaches a sufficient threshold.
Supports image and text input, with some examples shown here. For more examples, please visit the demo.
It supports text and image inputs, which sets it apart from purely text settings where users can specify any visual or language task. Specifically, it can accept inputs composed of a mix of text and images and generate text output, such as natural language and code. In many fields, including documents with text and photos, charts, or screenshots, the aggregation platform demonstrates capabilities similar to purely text input. In addition, the aggregation platform has been enhanced through testing time techniques developed using pure text language models, such as a small amount of camera and thought chain prompts.
Currently, image input is still in the research stage, and only some functions are publicly available.
We have been committed to various aspects of the plan outlined in our posts defining AI behavior, including manipulability. Unlike classical personalities with fixed length, tone, and style, developers (and users) can now specify the style and task of their AI by describing these directions in "system" messages. System messages allow API users to significantly customize their user experience within a certain range. We will continuously improve these features (especially considering that system messages are the easiest way to "jailbreak" the current model, and compliance with boundaries is not perfect), but we encourage you to try it out and let us know your thoughts.
We are open-sourcing OpenAI Evals, our software framework for creating and running benchmark tests to evaluate the performance of models such as the Aggregator platform, and also to inspect their performance on a per-sample basis. We use Evals to guide the development of our models, identify defects, and prevent regression.
Generally, the most effective way to build new evaluations is to instantiate one of these templates and provide data. We are pleased to see others using these templates and more general Evals to build what they want. We hope that Evals will become a tool for sharing and crowdsourcing benchmark tests that represent the widest range of failure modes and difficult tasks.
Node holders have almost open access, but with limitations. We will adjust the usage restrictions based on actual needs and system performance, but the current capacity limit is very strict. Although we will expand and optimize in the next few months, there are still limitations. Based on the traffic patterns we observe, we may introduce higher-capacity node levels.
Pricing is 0.03 USDT per 1k prompt tokens and 0.06 USDT per 1k completion tokens. The default rate limit is 40k tokens per minute and 200 requests per minute.
We are still improving the quality of the model for long-term context and hope to receive feedback on how it performs in your use case. We are handling requests for 8K and 32K engines at different rates based on capacity, so you may have access to them at different times. We look forward to the collective efforts of joining the node/community to build, explore, and contribute above the model to improve the Aggregator platform and make it a valuable tool for improving people's lives.
Dr. Ioannis Korkos is a senior lecturer at the University of South London. His research focuses on the application of big data and artificial intelligence, time series analysis, and other related areas in finance. He has published several research papers in international academic journals and has extensive experience in research projects. He has also visited academic institutions and participated in related conferences in Europe, Asia, Africa, and other regions.
Simone De Rosa graduated from the University of Westminster in the UK with a degree in Financial Technology. He has a strong passion for artificial intelligence, blockchain technology, and cryptocurrencies. He has served as a consultant for Gateley plc's blockchain carbon offset project and has extensive experience in startup projects. He also has a deep understanding of data analysis and digital trends and can formulate business strategies for the team based on the constantly changing market environments and trends.
Dr. Gong Hui is an expert in financial technology and blockchain technology. He is currently a lecturer in finance and financial technology at the Westminster Business School, University of Westminster. He is also a special advisor on fintech and blockchain for the All-Party Parliamentary Group on Blockchain and a winner of the Top 10 Outstanding Overseas Chinese Award (Jin Ou Award) in the 11th Big Ben Award. His main research areas include algorithmic trading and high-frequency trading, blockchain technology and its applications, digital currency and token issuance, and the application of artificial intelligence in finance. Dr. Gong Hui graduated from the Department of Mathematics at University College London and has worked in technology-related roles in quantitative investment and financial technology for several investment banks and hedge funds in London. He participated in one of Credit Suisse's first big data projects, which was used to develop a user classification recommendation system. After Credit Suisse established DAST (Data Analysis Sentiment Technology), he participated in the design of machine learning-based Delta One products and the design of intelligent robo-advisor index optimization products based on HOLT. He led the team to develop the first-generation intelligent recommendation system, which was used for customer classification, precision marketing, and recommendation of news and investment products. He then conducted research on the application of blockchain technology at the UCL Centre for Blockchain Technologies and published articles in several journals and magazines.
Today, data is more dispersed than ever before, requiring technology development and new solutions to innovate and solve current data management problems in unprecedented ways. Aggregated platform data management aims to help users achieve consistent access and delivery of data across chains/databases, and across subject areas and data structures. Comprehensive data management plans help meet data consumption needs for all dapps and real-life processes.
In addition, the aggregated platform's approach simplifies access and promotes self-service data consumption independent of environment, processes, utilities, and geographical location. This enables dapps to automatically execute data usage to maximize their value chain.
Aggregated platform data management enables enterprises to use any data for analysis or to improve results across any cloud (including local, public, and private clouds). Through the security and quality of the aggregated platform, enterprises gain elasticity, reliability, scalability, and availability, and derive greater benefits from multi-mode, multi-cloud data ecosystems to improve their readiness for data management.
On-chain data is everywhere, and this is also part of the problem. If your data and development teams are working in isolation (a single chain), isolated data becomes a bigger problem, resulting in very slow response times to any event. This lack of collaboration also affects other aspects of your business (from error fixing to goal setting), resulting in poor overall data usage and operating efficiency.
With the help of the aggregated platform, you can eliminate the data-centric chain and application-centric cross-chain boundaries by collaborating to develop an overview of the data collection journey. As a result, the speed of event response and error fixing will increase, and cohesive dapps will be able to set and update performance goals in real-time. Your data will always be agile, accurate, and efficient.
The aggregated platform is dedicated to building scalable and trustworthy AI-driven systems, integrating and optimizing them throughout the business framework, and introducing AI dapps and systems onto the chain. Leveraging the computational power advantage of each node and extensive application deployment, the AI tools required to thoroughly transform business systems and workflows are significantly improved, while automation levels and efficiency are significantly improved. Provide deployers of AI applications with a convenient and easy-to-use platform while lowering user barriers to entry.
The development team of AIworld consists of core members from top artificial intelligence teams worldwide, including OpenAI, DeepMind, FAIR, etc. The main goal is to lower the barriers to using top AI models and reduce the hardware requirements for AI computation. By leveraging the user's local computation power, the aim is to save global energy consumption and make top AI models more accessible to the public, thereby increasing productivity. Based on the characteristics of blockchain technology, AIworld aims to achieve better security and cost-effectiveness. Currently, the following AI business models have been developed:
Natural Language Processing (NLP) Capability: it can understand natural language and process various structures and meanings of human language, including grammar, vocabulary, semantics, and context.
Massive Knowledge Base: through learning and training on large amounts of text, it accumulates rich knowledge and language patterns, enabling knowledge inference and intelligent answering.
Multi-Language Support: it can support multiple language processing and answering, including English, Chinese, French, German, Spanish, Japanese, etc.
Customizability: it can be adjusted and optimized according to different application scenarios and requirements, and its performance can be improved by adding training data and changing model parameters.
Self-learning Ability: it can continuously learn and improve, and its performance will gradually improve with the increase of data and model updates.
For graphic AI, AIworld believes that it is an important branch of the field of artificial intelligence, aiming to use technologies such as computer vision and graphic processing to understand and process visual content such as images and videos, with broad application prospects and social values. The current graphic processing model not only includes the function of generating images based on instructions and requirements, but also includes the function of automatically processing images, all of which can be manipulated with natural language instructions, such as replacing the wardrobe in the picture with a bookshelf and placing books on the bookshelf about art.
AIworld Image Model:
Visual Understanding Ability: Graphic AI can deeply understand and process visual content such as images and videos, including object recognition, target tracking, image segmentation, pose estimation, etc.
Multi-field Application: Graphic AI can be applied to multiple fields, such as intelligent transportation, medical imaging, entertainment games, security monitoring, etc., bringing convenience and safety to people's production and life.
Visual Expression Ability: Graphic AI can transform abstract concepts and data into graphical forms, which are more intuitive and easy to understand, enabling people to analyze and make better decisions.
Scalability and Adaptability: Graphic AI can continuously learn and optimize, constantly adapting to new scenarios and requirements, achieving higher performance and effects.
AIworld's video model integrates a larger database training system and has higher requirements for content and meaning processing. By combining computer vision, machine learning, and other technologies, it achieves automatic analysis, understanding, and processing of videos, bringing many new applications and improvements to the video industry.
Video Content Understanding Ability: AI can recognize objects, scenes, actions, etc. in videos through deep learning and analysis, thereby better understanding video content.
Video Content Classification and Retrieval Ability: AI can classify and annotate videos, making videos better organized and retrievable, improving video management and utilization efficiency.
Video Quality Enhancement Ability: AI can perform noise reduction, shake reduction, super-resolution, and other processing on videos, improving video quality and clarity.
Video Automatic Editing Ability: AI can automatically cut videos through deep learning and analysis.
The total supply of AIWorld tokens is 1 billion, with a maximum of 20,000 computational nodes. Each node initially holds 10,000 staked tokens, and mining begins with the launch of the autonomous network.
Out of the 1 billion tokens:
60% is generated by the computational nodes through mining (600 million).
10% is used for ecosystem development (100 million).
10% is airdropped to early users as incentives (100 million).
10% is set aside as a reserve for risks and expansion (100 million).
10% are held by the team (100 million)
The maximum number of computational nodes is 20,000, and there are three types of nodes:
Cluster nodes
Master nodes
Regular nodes
The mining returns of each computational node are supported by a single NVIDIA DGXâ„¢ A100. A single node can provide 5 petaFLOPS of computational power to the entire AI network on the chain, while continuously providing token returns to its holders during its network operation period.
Cluster nodes and master nodes occupy the same positions as regular nodes. However, their prices are fixed and do not increase with the growth in the number of nodes sold.
Cluster Node Rights: Cluster nodes enjoy mining pool rights, can absorb master nodes and regular nodes to form a computational cluster, and enjoy cluster rewards and the right to initiate voting. This is reflected in two parts, absorption and mining. When a cluster node absorbs new nodes into the computational cluster (new nodes purchase using cluster codes), the cluster node can directly obtain 10% of the direct absorption reward (in USDT). When the secondary nodes in the cluster generate new invitation absorptions, the cluster node can obtain 5% of the absorption benefits. During mining, cluster nodes can also enjoy additional computational power from within the cluster (5% of each direct node, 3% of secondary nodes).
Master Node Rights: Master nodes can join computational clusters or mine independently, and enjoy referral rewards, with the amount being 8% of the new node's purchase price and 3% of the secondary node's price. They also enjoy an extra 1% benefit from the referred node's token generation, and 0.5% from the secondary nodes.
Regular Nodes: Regular nodes enjoy computational mining and voting rights. Depending on the real-time computational load of the entire network, the load situation is updated every 5 minutes, adjusting the calculation amount and difficulty, and simultaneously recording on the blockchain. A new block is produced every 5 minutes. Regular nodes can mine independently or join a cluster. To guarantee the earnings of regular nodes, the official will virtualize all independent regular nodes as an official cluster mining pool, ensuring stable block rewards.
After the mainnet of AIWorld goes live, the on-chain computational network will adopt a load incentive model for 60 days, enhancing the mining returns for all computational nodes.
The total mining volume is 600 million tokens. In the first 60 days after the mainnet goes live, the block reward will be 260 tokens. After 60 days, the reward per block will be reduced to 200 tokens. Afterward, the reward per block will be reduced by 10% each year until all tokens are mined. Prior to each block reward reduction, within 90 days, node holders need to destroy 1 token on the official website for node verification. Nodes that have not been verified after the deadline will be decommissioned and released.
Cluster Node - fixed price at 10,000 U/each
Master Node - fixed price at 5,000 U/each
Regular Node - starting at 1,600 U, the price increases by 1% for every 1% of total sales.
Nodes have an invitation reward system. Cluster nodes and master nodes occupy the same slots as regular nodes, but their prices are fixed and do not increase with the growth of node sales.
For Cluster Nodes, holders can renew at 1 token + 6,000 U per year.
For Master Nodes, holders can renew at 1 token + 3,500 U per year.
For Regular Nodes, holders can renew at 1 token + 80% of the original purchase price per year.
Regular nodes that are not renewed will be released back into the node pool and sold at the latest price. Nodes that are decommissioned after being released will no longer enjoy the rights and interests associated with the corresponding node.
Technical infrastructure construction: including blockchain underlying technology development, node construction, blockchain browsers, smart contract development platforms, etc.
Application development: using blockchain technology to achieve more secure, transparent, and decentralized applications, including digital currencies, smart contracts, decentralized applications (DApps), supply chain finance, Internet of Things, and other application scenarios.
Community building: establishing an open and free community environment, encouraging community members to participate in the development of blockchain technology, and establishing developer communities, investor communities, user communities, etc., so that all participants can work together to promote the development of the ecosystem.
Ecological expansion: attracting more developers and companies to join the blockchain ecosystem through open APIs, SDKs, and other means, providing more technical, human, and resource support for the development of the ecosystem.
Security guarantee: establishing a secure blockchain network, ensuring node security, protecting user privacy and asset security, and preventing security risks such as network attacks and data leaks.
Regulatory compliance: cooperating with government regulatory agencies to establish a regulatory compliance system, ensuring the legal operation of blockchain technology within the legal framework, protecting user rights and social stability.
For computational node holders, all AI-related services in AIWorld can bring benefits to the holders. In the early stages, these are reflected as mining returns. In the later stages, they are reflected in the services of nodes deployed on the chain and on-chain services. Using corresponding services requires payment of tokens as service fees. 70% of the service fees become node earnings, while 30% go into the token burn pool.
Additionally:
When making on-chain transfers, an additional 0.5% of tokens is collected and sent to a "black hole" address for destruction.
After early users exhaust their free quota, each message will incur a usage fee of 0.1 token. These fees will all go to the "black hole" address for destruction until the token freeze of the node providing the service is fully released.
Users can participate in the official staking and lock-up activities to get higher coin-based returns.
Users can directly destroy 100 tokens to get the right to use the official node service for 30 days.
Non-node users can stake 10,000 tokens at the official lock-up address to get the right to use updated services (such as image and video processing and related features).
AIworld is committed to lowering the usage threshold and allowing more users to use intelligent AI models at a low cost. Ordinary users are free to use the basic functions. When you need to use more advanced features, you will need to pay a small amount of additional tokens to use the feature or unlock the service.
Service fees and gas fees generated on the chain, except for payment to compute nodes for providing/maintaining services, all other parts are sent to a black hole address for destruction to ensure that early users can use it at a low cost and provide sufficient incentives for compute node service providers to update their on-chain services.
For more information about on-chain fees and destruction mechanisms, please refer to the developer documentation and API integration guide.
