AI Bible: Trans-AI: Transdisciplinary, Transformative and Translational: World Embeddings

Azamat Abdoullaev

THOU SHALT NOT MAKE A MACHINE IN THE LIKENESS OF THE HUMAN MIND (Dune, F. Herbert)

“It's quite conceivable that humanity is just a passing phase in the evolution of intelligence” (Geoffrey Hinton)

Trans-AI Commandments

THOU SHALT NOT MAKE A MACHINE IN THE LIKENESS OF HUMAN BODY

THOU SHALT NOT MAKE A MACHINE IN THE LIKENESS OF THE HUMAN BRAIN

THOU SHALT NOT MAKE A MACHINE IN THE LIKENESS OF THE HUMAN BRAINS

THOU SHALT NOT MAKE A MACHINE IN THE LIKENESS OF THE HUMAN BUSINESS

THOU SHALT NOT MAKE A MACHINE IN THE LIKENESS OF THE HUMAN

[AI BIBLE: the Old AI and The New AI; AI Bible: True AI and Real ML for Decision-Makers and C-Level Executives]

Trans-AI Commentary

Human-Imitating Artificial Intelligence (HAI) with Artificial General Intelligence (AGI), as generative AI, Deep Neural Networks, LLMs and ChatGPT-4, is the most hyped up deepfake news on the planet. The core idea of all this is a machine and/or computer, software and/or hardware, which perceives and learns, thinks and acts just like you, as if learning all by itself, without any human intervention and intelligence.

HAI/ML/DL/LLM applications outperforms humans at whatever specific task.

They exceed humans in big data processing, fast decision-making, rote learning, computing numbers, solving equations, memorizing, gaming, translation, recognition, search, driving, flying, running, or traveling underwater. They beat us at algorithmic trading and strategic games, make art and music or diagnosis, transcribe music, etc.

One should not be a great visioner to forecast that a general-purpose HAI as more efficient and productive human-like and human-level autonomous automated entities will replace humans at all human functions and jobs, tasks and positions, thus inventing the Simulated Hyperreality of AI-generated Simulacra Humans.

Trans-AI as World Embeddings

All Machine Learning/AI/DL/Large Language Models work with numerical data, vector spaces. Before any operation, all text/image/audio/video data has to be transformed into numerical representations. Embeddings are data that has been transformed into n-dimensional matrices for deep learning NN computations. Embeddings represent real-world objects, like words, images, or videos to be digitally processed.

Embeddings are foundational for AI, being vectors or arrays of numbers that represent the meaning and the context of tokens processed and generated by the model. Embeddings enable the AI model to handle multimodal tasks, such as image and code generation, by converting different types of data into a common representation.

Embeddings are an essential component of the transformer architecture that GPT-based foundation LLM models use, and they can vary in size and dimension depending on the model and the task. It is used to encode and decode the input and output data, for text classification, summarization, translation, generation, image generation or code generation.

Embeddings created can generalize to other tasks and domains through transfer learning — the ability to switch contexts — the reason of its popularity across AL/ML applications developers [See Popular Embedding Models]

As vector representations of data, embeddings capture meaningful, semantic or syntactic, relationships between entities, which could be thoughts or concepts, words or things, with all possible interactions, interdependencies and connections.

But today's AI embeddings rely on Rote Memory and Pattern Matching, having No Analogy, No Deduction, but simple Mathematical Induction or Statistical Inference from statistical patterns and correlations.

Trans-AI is based on the World Embeddings, the representations or encodings of the world or reality and its content and domains, where its

Entity Embeddings (EEs) are vector representations of categorical or ordinal or interval or ratio or cardinal variables or entities in a dataset. The EEs are learned by training a neural network to capture the relationships between different entities in a high-dimensional space of the encoded world or its domains, systems or processes.

Entity embeddings convert Entity Data into continuous numerical data, enabling the use of AI/ML algorithms that require numerical inputs.

Trans-AI Embeddings are covering the word embeddings, the representations or encodings of tokens, such as sentences, paragraphs, or documents, in a high-dimensional vector space, where each dimension corresponds to a learned feature or attribute of the language, as in LLMs.

The World Embeddings enables the AI/ML/LLM models to effective, efficiently and sustainably interact with the world, understanding semantic and syntactic relationships between the tokens to generate intelligible content.

Trans-AI as a Hyperintelligent Complement to Human Intelligence

Transformative, transdisciplinary and translational research has attracted significant and increasing interest in science and engineering. In scientific history, transdisciplinary and translational approaches have fostered new perspectives, areas, techniques, and results. Typical areas are sustainability research, translational public health and translational medicine, biomedical research, and transformative digitalization. Trans-AI/DS encourage big and outside-the-box thinking beyond the classic AI, data-driven, model-based, statistical, shallow and deep learning hypotheses, methodologies and developments.

Transdisciplinary/Transformational/Translational AI is about building a general-purpose, real, and true AI technology, an integration of different AI/ML/DL/NLP techniques, to complete humans at non-human innovative jobs and hyperintelligent tasks.

TransAI is to embrace all the valuable special AI innovations, ML techniques and LLM models, Data Science technologies and DL algorithms, like as mentioned in the Gartner Hype Cycles for Emerging Technology and AI, as sampled below.

Data-centric AI: Innovations in data-centric AI include synthetic data, knowledge graphs, data labeling and annotation.

Model-centric AI

Despite the shift to a data-centric approach, AI models still need attention to ensure the outputs continue to help us to take better actions. Innovations here include physics-informed AI, composite AI, causal AI, generative AI, foundation models and deep learning.

Composite AI refers to the fusion of different AI techniques to improve the efficiency of learning and broaden the level of knowledge representations. Since no single AI technique is a silver bullet, composite AI ultimately provides a platform to solve a wider range of business problems in a more effective manner.

Causal AI includes different techniques, like causal graphs and simulation, that help uncover causal relationships to improve decision making. Causal AI identifies and utilizes cause-and-effect relationships to go beyond correlation-based predictive models and toward AI systems that can prescribe actions more effectively and act more autonomously.

Applications-centric AI: Innovations here include AI engineering, decision intelligence, operational AI systems, ModelOps, AI cloud services, smart robots, natural language processing (NLP), autonomous vehicles, intelligent applications and computer vision.

Human-centric AI: This group of innovations includes AI trust, risk and security management (TRiSM), responsible AI, digital ethics, and AI maker and teaching kits.

Generative AI

The critical technologies that fall into the GenAI category include the following:

• Artificial general intelligence (AGI) is the intelligence of a machine that can accomplish any intellectual task that a human can perform.
• AI engineering is foundational for enterprise delivery of AI solutions at scale. The discipline creates coherent enterprise development, delivery, and operational AI-based systems.
• Autonomic systems are self-managing physical or software systems performing domain-bounded tasks that exhibit three fundamental characteristics: autonomy, learning and agency.
• Cloud AI services provide AI model building tools, APIs for prebuilt services and associated middleware that enable the building/training, deployment and consumption of machine learning (ML) models running on prebuilt infrastructure as cloud services.
• Composite AI refers to the combined application (or fusion) of different AI techniques to improve the efficiency of learning to broaden the level of knowledge representations. It solves a wider range of business problems in a more effective manner.
• Computer vision is a set of technologies that involves capturing, processing and analyzing real-world images and videos to extract meaningful, contextual information from the physical world.
• Data-centric AI is an approach that focuses on enhancing and enriching training data to drive better AI outcomes. Data-centric AI also addresses data quality, privacy and scalability.
• Edge AI refers to the use of AI techniques embedded in non-IT products, IoT endpoints, gateways and edge servers. It spans use cases for consumer, commercial and industrial applications, such as autonomous vehicles, enhanced capabilities of medical diagnostics and streaming video analytics.
• Intelligent applications utilize learned adaptation to respond autonomously to people and machines.
• Model operationalization (ModelOps) is primarily focused on the end-to-end governance and life cycle management of advanced analytics, AI and decision models.
• Operational AI systems (OAISys) enable orchestration, automation and scaling of production-ready and enterprise-grade AI, comprising ML, DNNs and Generative AI.
• Prompt engineering is the discipline of providing inputs, in the form of text or images, to generative AI models to specify and confine the set of responses the model can produce.
• Smart robots are AI-powered, often mobile, machines designed to autonomously execute one or more physical tasks.
• Synthetic data is a class of data that is artificially generated rather than obtained from direct observations of the real world.

• AI simulation is the combined application of AI and simulation technologies to jointly develop AI agents and the simulated environments in which they can be trained, tested and sometimes deployed.
• AI trust, risk and security management (AI TRiSM) ensures AI model governance, trustworthiness, fairness, reliability, robustness, efficacy and data protection.
• Causal AI identifies and utilizes cause-and-effect relationships to go beyond correlation-based predictive models and toward AI systems that can prescribe actions more effectively and act more autonomously.
• Data labeling and annotation (DL&A) is a process where data assets are further classified, segmented, annotated and augmented to enrich data for better analytics and AI projects.
• First-principles AI (FPAI) (aka physics-informed AI) incorporates physical and analog principles, governing laws and domain knowledge into AI models. FPAI extends AI engineering to complex system engineering and model-based systems
• Foundation models are large-parameter models trained on a broad gamut of datasets in a self-supervised manner.
• Knowledge graphs are machine-readable representations of the physical and digital worlds. They include entities (people, companies, digital assets) and their relationships, which adhere to a graph data model.
• Multiagent systems (MAS) is a type of AI system composed of multiple, independent (but interactive) agents, each capable of perceiving their environment and taking actions. Agents can be AI models, software programs, robots and other computational entities.
• Neurosymbolic AI is a form of composite AI that combines machine learning methods and symbolic systems to create more robust and trustworthy AI models. It provides a reasoning infrastructure for solving a wider range of business problems more effectively.
• Responsible AI is an umbrella term for aspects of making appropriate business and ethical choices when adopting AI. It encompasses organizational responsibilities and practices that ensure positive, accountable, and ethical AI development and operation.

AI Bible: True/Nonhuman/Real AI vs. False/Human/Simulacra AI

Interactive AI (IAI) or Agent Intellect: ML > DL > ANI > GenAI > AGI > ASI > Active Intelligence = Trans-AI = Meta-AI

Transdisciplinary AI: Machine Knowledge as a Super Scientist

Trans-AI/DS: transformative, transdisciplinary and translational artificial intelligence and data science

After the many ups and downs over the past 70 years of AI and 50 years of data science (DS), AI/DS have migrated into their new age. This new-generation AI/DS build on the consilience and universology of science, technology and engineering. In particular, it synergizes AI and data science, inspiring Trans-AI/DS (i.e., Trans-AI, Trans-DS and their hybridization) thinking, vision, paradigms, approaches and practices. Trans-AI/DS feature their transformative (or transformational), transdisciplinary, and translational AI/DS in terms of thinking, paradigms, methodologies, technologies, engineering, and practices. Here, we discuss these important paradigm shifts and directions. Trans-AI/DS encourage big and outside-the-box thinking beyond the classic AI, data-driven, model-based, statistical, shallow and deep learning hypotheses, methodologies and developments. They pursue foundational and original AI/DS thinking, theories and practices from the essence of intelligences and complexities inherent in humans, nature, society, and their creations.

Pursuing Trans-AI/DS research requires disruptive, outside-the-box, and ‘beyond’ thinking. Examples of beyond thinking for Trans-AI/DS include:

• beyond hypothesis,
• beyond data-driven,
• beyond model-driven,
• beyond statistical i.i.d. assumptions, and
• beyond the fitting approach.

Trans-AI: How to Build True AI or Real Machine Intelligence and Learning

We are at the edge of colossal changes. This is a critical moment of historical choice and opportunity. It could be the best 5 years ahead of us that we have ever had in human history or one of the worst, because we have all the power, technology and knowledge to create the most fundamental general-purpose technology (GPT), which could completely upend the whole human history.

The most important GPTs were fire, the wheel, language, writing, the printing press, the steam engine, electric power, information and telecommunications technology, all to be topped by real artificial intelligence technology.

Our study refers to Why and How the Real Machine Intelligence or Transdisciplinary/Transformational/Translational AI (Trans-AI) or Real Superintelligence (RSI) could be designed and developed, deployed and distributed in the next 5 years.

The whole idea of RSI took about three decades in three phases.

The first conceptual model of TransAI was published in 1989. It covered all possible physical phenomena, effects and processes.

The more extended model of Real AI was developed in 1999. A complete theory of superintelligence, with its reality model, global knowledge base, NL programing language, and master algorithm, was presented in 2008.

The RSI project has been finally completed in 2020, with some key findings and discoveries being published on the EU AI Alliance/Futurium site in 20+ articles. The RSI features a unifying World Metamodel (Global Ontology), with a General Intelligence Framework (Master Algorithm), Standard Data Type Hierarchy, NL Programming Language, to effectively interact with the world by intelligent processing of its data, from the web data to the real-world data. The basic results with technical specifications, classifications, formulas, algorithms, designs and patterns, were kept as a trade secret and documented as the Corporate Confidential Report: How to Engineer Man-Machine Superintelligence 2025. As a member of EU AI Alliance, the author has proposed the Man-Machine RSI Platform as a key part of Transnational EU-Russia Project. To shape a smart and sustainable future, the world should invest into the RSI Science and Technology, for the Trans-AI paradigm is the way to an inclusive, instrumented, interconnected and intelligent world.