The Man-Machine Interaction Ontology: The Reality AI Engine: Causal Machine Intelligence and Learning

The Man-Machine Interaction Ontology: The Reality AI Engine: Causal Machine Intelligence and Learning

"Everything is Interaction and Reciprocal", Humboldt

"Reality appears as a dynamically interdependent process. All factors, mental and physical, subsist in a web of mutual causal interaction, with no element or essence held to be immutable or autonomous", Macy

I argue for the Man-Machine Metaphysics/Ontology/Science/Engineering which is after the consistent and complete, systemic and systematic scientific world model featuring the fundamental [onto-scientific-engineering] principle:

All Reality is Interaction, or "Everything is Interaction and Reciprocal", and "Reality appears as a dynamically interdependent process. All factors, mental and physical, subsist in a web of mutual causal interaction, with no element or essence held to be immutable or autonomous".

It is designated as "the AA Interaction Principle of the Universe", implying three major facts and propositions:

The World, Reality, Being, Existence, or Universe is the Sum Total of All Interactions.

All of reality is interaction; interactions create all the substances, states, changes and relationships, all the networks and systems, all the phenomena and processes, forces and emerging properties,?including all the intelligence, natural or artificial.

Everything interacts with everything else: something (A/X) causes something else (B/Y) if and only if the something else (B/Y) causes the something (A/X).

Corollaries:

All interacts; nothing exists in isolation without interactions.

Something exists and changes, if it interacts with something else, having an effect on each other.

Anything is a node of interactions, being a net of interactions with the world around it.

Any intelligence consists in causal learning, inference and understanding to effectively interact with the world.

Any real intelligence, human or machine, deals with reality in terms of the world models and data/information/knowledge representations for cognition and reasoning, understanding and learning, problem-solving, predictions and decision-making, and interacting with the environment.

Real AI Machine Intelligence must have the world model learning, intelligence and inference engine to meaningfully and effectively or causally interact with the world, including nature, machines, humans, the internet, and other real-world networks.

AA/AI Rule: "Without understanding the cause and effect of interactions within the world, no AI model, algorithm, technique, application, or technology is real and true", be it:

Natural language generation converting structured data into the native language

Speech recognition converting human speech into a useful and understandable format by computers

Virtual agents, computer applications that interact with humans to answer their queries, from Google Assistant to the IMB Watson

Biometrics

Decision management systems for data conversion and interpretation into predictive models

Machine learning empowering machine to make sense from data sets without being actually programmed, to make informed decisions with data analytics and statistical models

Robotic process automation configuring a robot (software application) to interpret, communicate and analyze data

Peer-to-peer network connecting between different systems and computers for data sharing without the data transmitting via server

Deep learning platforms based on ANNs teaching computers and machines to learn by example just the way humans do

Generative AI models with image generation algorithms

AI optimized hardware support artificial intelligence models, as neural networks, deep learning, and computer vision, including CPUs, GPUs, TPUs, OPUs to handle scalable workloads, special purpose built-in silicon for neural networks, neuromorphic chips, etc.

Real AI is NOT about representing computational models of intelligence, described as structures, models, and operational functions that can be programmed for problem-solving, inferences, language processing, etc.

Real AI is about the computational models of reality and mentality, described as causal structures, models, and operational functions that can be programmed for problem-solving and inferences for a wide range of goals in a wide range of environments.

So, the iron rule of real AI: without understanding the cause and effect of interactions within the world, there is no True, Autonomous Machine Intelligence

The Interaction Mechanism of Causation and Causality

We might divide the Humanity Intelligence Development Era into two Epochs:

The One-Way, Direct and Linear Causality Epoch till the 21st century:

"X causes Y or Y causes X"

The Mutual and Reversed, Nonlinear Causality Interaction Epoch since the 21st century:

"X causes Y if and only if Y causes X"

[Causal] Interaction is the centerpiece of the universe and so the main subject of?all sciences and technologies and practices. Comprehending the nature, meaning, kinds, varieties, and ordering of interactions amounts to knowing the beginnings and endings of things, to uncovering the implicit mechanisms of world dynamics, or to having the fundamental scientific knowledge.

“Alles ist Wechselwirkung”, or " All is Interaction", as wrote Alexander von Humboldt, who in the early 19th century the General Physics of the Earth, the foundations for today's Earth system sciences.

Crucially, the Interaction World Paradigm enables creating the Universal Machine Intelligence System featured by the General AI World Model driven by the Reality AI Engine. It is the only Path Towards Autonomous Machine Intelligence with the world model/knowing/learning/inference/interaction architecture.

Our sustainable future depends on emerging technologies led by real artificial intelligence and causal machine learning, which is intelligent robotics and automation, virtual, augmented and extended reality, the internet of things, quantum computing, bioscience, semiconductors and autonomous systems.

By 2025, the working prototype as a real machine intelligence minimum valuable product (MVP) is intended to be created.

Towards the Universal AI World Model Learning and Inference Engine ??

Humanity is involved in its last socio-technological revolution, the AI revolution, which fundamentally changes the ways people live, play, work or study. As such, emerging AI technology is supported by three major aspects:?world model, domain knowledge, data processing, and causal machine learning which is autonomously discovering the causal patterns and laws from the world's data, all to effectively interact with the world.???

My goal is to create the machine world model as the intelligence framework for universal AI world model learning and inference engine, identifying its key variables, prime entities, processes and interactions, as its major categories, classes, concepts and interrelationships.

Below it is established, basing on the vast amounts of evidence, data and scientific knowledge, that interaction is the only real and natural relationship, with all the rest as its derivatives, be it causal relations, temporal relations, spatial relations, whole-part relations, and what else.

In all, the world is the universal system of interactions, the global network of interacting networks. The real-world interrelationship of interaction is all and everything what is sufficient and necessary to understand and adapt, navigate and interact with the world of reality.

My final goal is not to innovate "interaction" to the human world of knowledge, what might take several lives considering our cognitive biases and social prejudices and strict loyalty to our schools of thought and in-group thinking.

The objective is rather to formalized an interactive world model for intelligent machines, thus making free of all human cognitive biases and social prejudices and trained presumptions coming from the strict loyalty to our schools of thought and in-group thinking.

Specifically, I developed the Universal Machine Intelligence Model (UMIM), as real autonomous artificial intelligence (RAAI) with deep causal learning, knowledge and understanding about the world of reality provided by its key components/modules:

Machine World Model;

Master Algorithm;

World Data Framework World Data;

Global Knowledge Base World.Net;

Domain Knowledge Base Domain.Net

The core of UMIM AI systems is a world model engine (WME) that effectively interacts with the world, knowing its states, changes and trends. It is highly likely that animals and people each have their own "world model", as some believe somewhere in their prefrontal cortex, like Y. LeCun has suggested in "A Path Towards Autonomous Machine Intelligence", Version 0.9.2, 2022-06-27.

That the world model is a make or break point could be proved by the recent Meta's failing with Galactica. Galactica is a large language model for science, trained on 48 million examples of scientific articles, websites, textbooks, lecture notes, and encyclopedias. Meta promoted its model as a shortcut for researchers and students. In the company’s words, Galactica “can summarize academic papers, solve math problems, generate Wiki articles, write scientific code, annotate molecules and proteins, and more.”

Like all language models without any world modeling, Galactica is a mindless statistical bot that cannot tell fact from fiction, generating biased and incorrect results on social media.?

Due to being statistical nonsense on the Meta scale, it survived?only 3 days.

To have the AI world model, it is necessary to describe, formalize and encode its basic categories, classes and types as the fundamental modes of existence/being/reality/world/universe, with its universal class of reality.

From our rich experience and practice, it is plain and clear that one of the essences of the world is interaction or interacting with each other, with machines, with our environment and the whole reachable world.

This is due to the fundamental fact that interaction is a reciprocal action and influence, or complete reversed causation, creating all real-world systems and complex networks.

This implies that interaction or interacting or causal interaction or interactive causation is the centerpiece of the universe; for nothing could emerges, exists, changes or ceases to exists without interacting with its environment, as the environment itself.

It is everywhere and beyond as the most fundamental modes of existence, but we know practically nothing about it, its structure and mechanism, input and output, scope and influence. Interaction is critical at all levels of reality, making or breaking its most complex structures or simple systems, from the atom to the whole world, as the global network of all interacting networks.

But try and Bing and Google the "Interaction" keyword to retrieve some poor results to read or study, if only the fresh wiki article "Interaction" under construction, while having totally 3+bn results.

Oxford defines the term "interaction" as "reciprocal action or influence".

Physics, a particular way in which matter, fields, and atomic and subatomic particles affect one another, e.g. through gravitation or electromagnetism.

Some dictionaries define it as "a mutual or reciprocal action or influence".?

As its subordinates go interplay, interchange, interconnection, interlinkage, intercourse, association, relations or interrelation, interactivity and interconnectivity.

For example, interactivity is related to?interaction?between users and computers and other machines through a?user interface. Interactivity can however also refer to interaction between people. It nevertheless usually refers to interaction between people and computers – and sometimes to interaction between computers – through software, hardware, and networks. The interconnectivity deals with the interactions of interactions within systems, so that combinations of many simple interactions can lead to emergent phenomena.

Both metaphysics and science and statistics are generally ignoring its fundamental properties, what its nature, structure, mechanism, or types, and how interaction relates to causation and causality and interrelationship in general.

We know about "interactionism" as theoretical perspectives for explaining how the mind-brain interacting or how "social behavior is an interactive product of the individual and the situation".

In?physics, interactions viewed as the?fundamental interactions, fundamental forces governing the universe as:?

the?gravitational?and?electromagnetic?interactions, which produce significant long-range forces whose effects can be seen directly in everyday life,

the?strong?and?weak interactions, which produce forces at?subatomic distances?and govern nuclear interactions. Some scientists hypothesize that a?fifth force?might exist, and this prime force is Interaction itself as the real-world category with the real world causal power.

Fundamentally, the interaction is a deep unknown known unless we discover its working mechanism and how it completes causality and causation as the reversed causality and the reciprocal causation.

Interaction is All and Everything

Interaction occurs as two or more things or or entities or objects have a causal, reciprocal and mutual effect upon one another, like as additive/positive interactions, systemic/synergistic/reinforcing interactions or antagonistic/negative interactions. The idea of a two-way nonlinear effect is essential in the real-world interelationship of interaction, as opposed to a one-way linear causal effect.

Then real-world causation or real-life causality should be scientifically defined in terms of interactions among processes, events, becoming, happenings, all described as the causal factors or variables or data.

Or, a causal interaction (depending on the nature of the interaction, it might also be called a causal force) is a process by which?entities?interact with each other.

Whenever things are connected by causal interactions, they form causal interaction networks that are generally classified by the nature of the entities involved, and dubbed as?interactome or connectome.

Interaction has different meanings in various sciences, but as causal interaction networks could be analyzed by graph theory, describing the structure, topology, functional modulus and different metrics, as degree distribution, and network science, considering distinct interacting elements, phenomena or actors represented by?nodes?(or?vertices) and the connections between the elements or actors as?links?(or?edges)..

There are different types and forms of interactions depending on the domain of actions:

  • Physical interactions between physical objects, from elementary particles to the cosmological structures; there are four known fundamental interactions in nature: the electromagnetic, strong, weak and gravitational interactions. The weak and electromagnetic interactions are unified in electroweak theory, which could in turn be unified with the strong force in a Grand Unified Theory and further with gravity in a theory of everything, but experimental results are yet to prove this.
  • Chemical interactions between atoms and molecules, embracing all chemical reactions
  • Biological interactions, as interactomes, biological networks, molecular interactions, protein-protein interactions, gene-environment interactions, cell-cell or gene-gene interactions, noninteractive, synthetic, asynthetic, suppressive, epistatic, conditional, additive, single-nonmonotonic and double-nonmonotonic; bio-medical drug interactions
  • Brain-Mind interactions, as the brain-mind interactions or neural interactions, neural networks, connectome?(a comprehensive map of neural connections in the brain), thought-action interactions or brain-machine interactions
  • Knowledge/Information/Data Interactions, all the science, engineering, the arts and technology are interconnected, interdependent and interrelated. No knowledge fields or disciplines exists in data/information/knowledge silos

No alt text provided for this image

  • Data interactions, Data obtained by interaction detection methods as stored in databases or ML models and available for computational analysis, like as protein-protein data interactions or interactions datasets that store event type data, such as?click?and?watch?event types, and event value data for each of your events
  • Social interactions, as exchange, competition, conflict, cooperation, and accommodation
  • Economic interactions between economic entities, consumers, businesses, institutions, organizations, markets, etc.
  • Political interactions between political parties, voters, institutions, etc.
  • Ecological interactions between ecosystems, food webs, species interactions, from neutralism to mutualism; climate networks
  • Technological interactions, technological networks, digital interconnections
  • Cybernetic interactions, causal feedback, positive and negative
  • Machine-Machine interactions, computer-computer communications, Digital interactions, the internet, as the interacting network of computing networks
  • ML Computing Interactions, ML/DL/ANNs, Deep Neural Networks algorithms, reinforcement learning, the error backpropagation algorithm reverse action, as the deep neural networks are struggling to perform,?applying the error/loss/cost/reward/profit/utility/fitness objective optimization function
  • Human-machine interactions, Web 1.0, a static information provider where people read websites but rarely interacted with them, Web 2.0, an interactive and social web enabling collaboration between users, Web 3.0 will change both how websites are made and how people interact with them
  • Globalization 1.0, the process of interaction and integration among the world's economies, cultures, populations, companies, and governments worldwide caused by the cross-border trade in goods and services, technology, and flows of investment, people, and information.
  • Globalization 2.0. Global Interactions of countries and cities, people and machines, ideas, cultures, religions, humans and nature; the Global Network, Global interconnections of different parts of the world resulting in the expansion of international social, cultural, economic, and political activities via the international institutions, interacting technologies and digital networks, all interrelated by global AI platforms.
  • AI-world interactions, AI-Human Interactions, AI-AI Interactions

To exists is to interact, to interact is to exist, or to be is to interact, to interact is to be.

If parallel universes really exist, they must interact, and “such an interaction could explain everything that is bizarre about quantum mechanics.”

No alt text provided for this image

The Interaction Mechanism of [Causation or Causality]: Causal Interaction as Global Process

We don't know that Causal Interaction is the only natural relationships and globally distributed process, the fundamental mechanism of existence, being and becoming, of all the world of reality.

Generally, we see causality and interaction as two distinct phenomena, badly missing Causal Interaction or Interactive Causality. It is true if not to mention the Principle of Causal Interaction for Mental Causation, known as Cartesian Interactionism:?some mental events interact causally with physical events.

We are shy to recognize three critical forms of interaction:

reverse causality, if only as "a possible explanation for associations between diabetes and certain types of cancer".

mutual causality, that there must be an interrelationship between two things, that mutual causality provides for an integrated understanding of our world as well as our life, explaining how the body handles and assimilates the many different influences that it encounters than can linear causality

causal interactions, for "reality appears as a dynamically interdependent process. All factors, mental and physical, subsist in a web of mutual causal interaction, with no element or essence held to be immutable or autonomous" (Macy 1991).

Instead, causality with its causal reasoning and inference is linearly interpreted in various terms of

empirical regularities,

changes in conditional probabilities,

counterfactual conditions,

mechanisms underlying causal relations,

a pattern of forces, causal power,

invariance under intervention.

A general metaphysical question about cause and effect is what kind of entity can be a cause and an effect. Cause and effect are of one and the same kind of entity, with causality a symmetric relation between them, as to the interactive causation.

That is to say, "A is the cause and B the effect" and/or "B is the cause and A the effect", as it is formulated by Bayes' theorem?(Bayes' law?or?Bayes' rule), which is actively used in probability theory and statistics and AI with ML/DL.

P (A/B) P (B) = P (B/A) P (A) = P (AB) = P (X, Y) = P(C,E)

where A/X and B/Y are events, or process, becoming, action, doing, change, or simply real-world variable.

P (AB) describes a statistical/probabilistic interaction among two or more variables, in which the effect of one causal variable on an outcome depends on the state of all other causal variables (that is, when effects of the two causes are not?additive or superposition or linear). As an example, the interaction nonlinear effect of education and ideology on concern about climate change will differ from the additive linear effect.

No alt text provided for this image

If the event of interest is?A?and the event?B?is known to have occurred (by assumption, presumption, assertion or evidence), "the conditional probability of?A?given?B", or "the probability of?A?under the condition?B", is written as?P(A/B). It can be reversed or converted according to the rule.

It describes several types of causal models developed as a result of observing causal relationships: common-cause-multi-effect relationships, common-effect-multi-cause relationships, causal chains, reversible causation, and causal?loops/cycles/homeostasis, causal relationships form a stable cycle or dynamic systems by reinforcing mechanism.

So, the introduction of interactions can have important implications for the interpretation of statistical models, statistical processes (regression analyses), classification algorithms and types of learning.

The real-world approach covers two main approaches in statistical classification, which are called the?generative?approach and the?discriminative?approach. They refer either to the three classes as?generative learning,?conditional learning, and?discriminative learning, or to?generative classifiers/algorithms?(joint distribution, P (X,Y)) and?discriminative classifiers/algorithms?(conditional distribution, P(Y/X) or no distribution). X represents observable variables, explanatory, independent variables or features, while Y - target, label or dependent variables. Both are a set of quantifiable properties which scaled as categorical, ordinal, integer-valued or real-valued variables or data.

The "classifier" refers to the mathematical?function, implemented by a classification algorithm, that maps input data to an output data of category. Since different data sets require specific algorithms, there is a large toolkit of classification algorithms:

Since no single form of classification is appropriate for all data sets, a large toolkit of classification algorithms have been developed:

In all, types of generative models are:

Discriminative models

Correlation Implies Causation

The interaction paradigm reviews many collective presumptions, like as "correlation does not imply causation", being a statistics mantra for many years: repeat after me, correlation is not causation, correlation is not causation, correlation is not causation

In fact, a correlation between X & Y can be explained as X causes Y, Y causes X, or some other variable causes both X and Y.

If Z causes both X and Y, but X doesn’t cause Y, and Y doesn’t cause X, then any correlation between X and Y would be spurious.

The key difference between experiments and correlational studies concerns the variable that is thought to be the cause.

In an experiment, this must be a manipulated variable – i.e., a variable that is completely under the control of the experimenter. In a correlational study, the cause is a measured variable, just like the effect. The manipulated variable in an experiment is called the “independent variable.” The measured variable in an experiment is called the “dependent variable.”

Both variables in a correlational study are measured, so we name them in terms what we think they are doing, instead of what they are. The one that we think is the cause (of the other) is often called the “predictor variable” and the one that we think is the effect (of the other) is often called the “predicted variable.”

A correlation between X & Y can be explained as: (1) X causes Y, (2) Y causes X, or (3) Z causes both X & Y. If you are interested in whether X causes Y, then the possibility that Y causes X, instead (which is called “reversed causation”), illustrates the “directionality problem”; while the possibility that some other variable, Z, causes both X & Y illustrates the “third-variable problem.”

Thus correlation implies causation as the key feature of interaction, which is causal interaction or complete, correlative causation.

All in all, our general analysis proves universal validity of the AA Interaction Principle of the Universe as the Interaction Mechanism of Causation implying that any [causal] action must have its retroaction or there is no action in the world without a reaction as mush as no reaction without action:

X causes Y if and only if Y causes X: X > Y iff Y > X

Or, causal interaction is a fundamental process by which?all the entities in the world?interact with each other, creating real-world systems and networks of various scope and complexity.

What is most crucial, causally interactive reality is the source model for all intelligence and learning, natural or artificial, human or machine, organic or non-organic, terrestrial or alien, specific or universal, narrow or general.

Reality AI vs. Human AI: the UMIM

The project of creating human-like and human-level artificial intelligence (hlhlAI) as Artificial General Intelligence started after World War II; for electronic computers could be both number-crunching machines and symbol-manipulating machines.

One could choose to pursue AI as machine intelligence and learning in three ways:

assuming that machine intelligence replicates/imitate only special parts of human intelligence, cognition or intelligent behavior, ML, DL, ANNs, developing artificial intelligence ONLY for specific purposes (ANI)

assuming that machine intelligence is NOT identical to human intelligence, Weak AI.

assuming that machine intelligence is identical to human intelligence, Strong AI.

My position is different, relying on science and its key subject of reality and interactions.

First, we SHOULD not develop computers with human-like intelligence, but with human-level intelligence and beyond.

Second, computer intelligence/power will never develop into human reason, the two are fundamentally different.

Third, computers SHOULD have a causal model of reality, or machine intelligence and learning (MIL) must be Causal/Real MIL

A single source of truth, knowledge and intelligence is reality and its interactive causation, with the fundamental causal laws, rules and patterns.

Causality/Interaction is a nuclear core of human thinking, particularly in science and technology. Nothing can be done without causes and effects.

Causal knowledge of the world is a decisive part of AI, and only computers handling causality can effectively interact with the world.

Again, AI mastering causality could deal with the critical causal questions, like if the climate change/global warming is caused by human activity (the release of greenhouse gases into the atmosphere) or it is just natural variations.

Acausal ANI, ML, DL is the application of mathematical methods as ANNs to huge amounts of data to find statistical correlations/associations/patterns and infer probabilities.

Causal AI, ML, DL is the application of ontological, mathematical and scientific methods to real-world data to find causal regularities/patterns/correlations/associations, infer predictions, prescriptions and create rational interactions.

What Is Real World AI Modeling?

The purpose of Real World AI models is to apply the world model engine to discover new patterns, predict outcomes or make decisions by understanding the interrelationships between multiple inputs of varying type to effectively interact with the world and its environment.

The creation of intelligent machine deep learning and inference models is the creation of Causal AI modeling that follows three basic steps:

  • Modeling:?The first step is to create a Causal AI world learning and inference model machine (Reality AI Engine), which uses a complex algorithm or layers of algorithms to interpret real-world data and make decisions based on that data.
  • AI model training:?The second step could be to train the Real AI Engine for special tasks or knowledge domains, processing large amounts of data through the Causal AI model machine in iterative test loops and checking the results to ensure accuracy, and that the model is behaving as expected and desired as it learns.
  • Inference:?The third step is machine discovery/reasoning/inference, from causal to categorical to analogical to realistic. This step refers to the deployment of the Causal AI world model into its real-world use case, where the RAI engine routinely infers real-world conclusions based on the data, information and knowledge.

Causal AI/ML/DL is a complex process with high computational, storage,?data security, and networking requirements. What could be supported by by AI hardware and software resources, like as Intel? Xeon? Scalable processors, Intel? storage and networking solutions, and Intel? AI toolkits and software optimizations, to design and deploy RAI/ML solutions with ease and cost efficiency.

Conclusion

All and everything interacts; for "Everything is Interaction and Reciprocal" (Humboldt). Here come the AA Universal Principle of the Universe, which is both metaphysical and physical, scientific and ontological.

For Real AI Machine Intelligence, it is an imperative to have the world model learning, intelligence and inference engine to meaningfully and effectively or causally interact with the world, including nature, machines, humans, the internet, and other real-world networks.

Sources

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

What Determines the World: Causality as the Life-or-Death Relationship

The centerpiece of the world: real causality/nonlinear causation: interrelationships and interactions: Real/causal AI/ML/DL

What is real intelligence? What is natural intelligence and artificial intelligence and how are they different from each other?

Interaction

MACY. J. (1991). Mutual causality in Buddhism and general systems theory. Albany, NY: State University of New York Press.

要查看或添加评论,请登录

Azamat Abdoullaev的更多文章

社区洞察

其他会员也浏览了