Apple, Google, Amazon, Microsoft, and Facebook, enjoying the highest ever $1+ trillion market cap, are still working on an old and obsolete statistics-based, weak and narrow, commercial AI/ML/DL that could potentially imitate some parts of the human brain, mind, or behavior.
While a true big goal is to create a real-world AI for the real-world complex problems, in all their reality and complexity, nonlinearity, dynamics and transdisciplinarity, which is transgressing as a Trans-AI of AI, ML and DL, a scientific, innovative, real AI vs. non-scientific, old-style, fake AI/ML/DL.
The Real SOTA of Real AI
Many “elite researchers in AI”, not mentioning the general public, are lullabying themselves that “human level machine intelligence,” or HLMI, has a 50 percent chance of occurring within 45 years and a 10 percent chance of occurring within 9 years.
We don’t see the general trend: machines are already surpassing humans in many domains. It is that narrow and weak AI applications outsmarting human minds in more and more fields:
Strategic gaming, like chess, the board game Go, and some Atari video games
Safety and security
Communication, as technology platforms, social media networks, bots and digital assistants
The existential question is when will AI be smart enough to outsmart people?
We one critical step away from a machine intelligence whose superintelligence transcendence/perfection largely depends on its causal power to detect, identify, process, compute, remember and manipulate any number of causal variables from any environment, physical, mental, digital, or virtual.
Now unknowingly by the general public, a Real-World AI is rising combining and transcending all the special designed intelligent algorithms.
We miss to see that the Real-World AI is around us, looking in the wrong direction of the big-tech companies, just fearing to be AI-disrupted, as Apple, Google, Amazon, Microsoft or Facebook, or big powers, as China, USA, or Russia.
Meantime, in 2020, a generally unknown i-company, EIS Encyclopedic Intelligent Systems LTD, has successfully completed its unprecedented R&D of the Transdisciplinary AI Model as a Real-World AI or a Causal Machine Intelligence and Learning, trademarked as Causal Artificial Superintelligence (CASI) GPT Platform complementing human intelligence, collective and individual.
The EIS Real-World Human-Machine Intelligence: the Trans-AI of AI/ML/DL
EIS Encyclopedic Intelligent Systems LTD (EU, Russia),being a member of EU AI Alliance, has successfully completed its strategic vision and mission of innovating the Transdisciplinary AI Model as a Real-World AI or a Causal Machine Intelligence and Learning.
It is trademarked as Causal Artificial Superintelligence (CASI) GPT Platform complementing human intelligence, collective and individual. The company has spent zero public funding and private investment for the outstanding discovery, which is set up to change the ways the world works, relying only on its own resources.
EIS is aimed to engineer a Real-World, Trans- AI, deploying it as a Global Human-Machine Internet Platform. It is open to the civic science platforms and large public and private investors from the EU, USA, China, Russia, UK, and socially responsible big tech companies, to develop Proof-of-Concept/Mechanism/Principle Prototype that can demonstrate the CASI feasibility for a full-scale global deployment.
The EIS Trans-AI Technology Platform serves as a unifying man-AI global industrial technology platform for all the key sectors. Here are some key areas that rely on the real-world artificial intelligence:
- Autonomous driving
- Agricultural industry
- Safety and security
- Communication, including technology platforms, social media networks, bots and digital assistants
- Health care
- Real estate
- Augmented reality and virtual reality
- Banking, trading, and other financial services
- Public administration
- Space exploration and astronomy
Real AI eats the world
Marc Andreessen famously said that “Software is eating the world”. And now we see how data is eating the world.
The point is everything surrounding us is data, generated by digital devices, internet of things or human beings.
We create a huge universe of data, a data hyperreality, which denizens are:
statistics, facts, recordings, observations, numbers,
data points, observations,
structured or unstructured data,
structured data in database rows,
unstructured data in text,
graphs, charts, plots, charts, statistic graphics,
text, data documents,
coded data, software, information, and knowledge.
By 2025, IDC predicts there will be 163 zettabytes of data. And all the hopes of dealing with the universe of ever-increasing data are desperately connected with data science and its new branches, as ML and DL.
For the majority of current successes of machine learning reduces to large scale pattern recognition on suitably collected independent and identically distributed (i.i.d.) data, encoding the general distribution of the problem into its parameters. However, in real-life scenarios, distributions can suddenly change, due to interventions in the world, domain shifts, temporal structure, etc. Examples include when convolutional neural networks trained on millions of images stop being able to see objects under new lighting, angles, or backgrounds.
Again, “generalizing well outside the i.i.d. setting requires learning not mere statistical associations between variables, but an underlying causal model.” “It is fair to say that much of the current practice (of solving i.i.d. benchmark problems) and most theoretical results (about generalization in i.i.d. settings) fail to tackle the hard open challenge of generalization across problems.” It comes from the researchers at the Montreal Institute for Learning Algorithms (Mila), Max Planck Institute for Intelligent Systems, and Google Research write in the study, titled “Towards Causal Representation Learning”. Who still mistake in selecting linear causal machine learning models, as “structural causal models” and “independent causal mechanisms” to allow the AI system to understand both the causal variables and their effects on the environment.
In fact, we need a Trans-AI model which is powerful to understand all massive data sets, as real world data and machine generated virtual data, their nature, data types, scales, levels, measurements, handling them with the speed of light.
It is not software or data, but the Trans-AI will eat the world.
The Trans-AI as the Most Sustainable Transdisciplinary Solutions of Global Threats
The Real-World AI is set up to change how the world works, being the engine of digital revolution, as well as all transdisciplinary science and technology, including community science and its projects.
The COVID-19 pandemic global crisis has accelerated the need for transdisciplinary solutions, of which one of the most disruptive innovations could be a Transdisciplinary AI (Trans-AI) or Real-World AI.
It is designed as the human-machine digital intelligent platforms facilitating integrated knowledge, competences, and workforce skills to meet massive technological unemployment. As mentioned, addressing societal challenges, as embedded in SDGs, using transdisciplinary research is considered a “mainstream modus operandi for research” by the OECD Global Science Forum (GSF).
In the AI and Robotics era, there is a high demand for the trans-disciplinary knowledge, competence, and high-technology training in a range of innovative areas of exponential technologies, such as artificial intelligence (AI), machine learning (ML) and robotics, data science and big data, cloud and edge computing, the Internet of Thing, 6G, cybersecurity and mixed reality.
The combined value – to society and industry – of digital transformation across industries could be greater than $100 trillion over the next 10 years. “Combinatorial” effects of AI and Robotics with mobile, cloud, sensors, and analytics among others – are accelerating progress exponentially, but the full potential will not be achieved without collaboration between humans and machines.
Given that, the Trans-AI is proposed to integrate disciplinary AIs, symbolic/logical or statistic/data, as ML Algorithms (Deep Learning (DL), Artificial Neural Networks (ANNs)), aiming to augment or substitute biological intelligence or intelligent actions with machine intelligence.
The Trans-AI is to be developed as a Man-Machine Global AI (GAI) Platform to integrate Human Intelligence with Narrow AI, ML, DL, Human-level AI, and Superhuman AI. It relies on fundamental scientific world’s knowledge, cybernetics, computer science, mathematics, statistics, data science, computing ontologies, robotics, psychology, linguistics, semantics, and philosophy.
Since it is widely recognized that the lack of reality with causality is the “root cause” of development problems of current machine learning systems, the Trans-AI is designed as a Causal Machine Intelligence and Learning Platform, to be served as Artificial Intelligence for Everybody and Everything, AI4EE.
The Trans-AI technology could make the most disruptive general-purpose technology of the 21st Century, given an effective transdisciplinary ecosystem of innovative business, government, policy-makers, NGOs, international organizations, civil society, academia, media and the arts. The Trans-AI Knowledge Graph covers ERC’s fields of research: Physical Sciences and Engineering (PE), Life Sciences (LS), and Social Sciences and Humanities (SH).
The Trans-AI is an advanced digital technology transdisciplinary project beyond discipline-specific approaches, involving all basic knowledge fields, as ontology, computer science, mathematics, statistics, data science, physics, cognitive sciences, psychology, linguistics, semantics, cybernetics, and general philosophy, among others, as well as an AI citizen science.
How the Trans-AI disrupts the Old-Style Human-Like AI
The Real-World, Trans-AI will eat the world, and its disruptive paradigm radically changes the standard concepts of AI, ML, and DL, among others, as defined here.
- Real AI is a set of knowledge models, techniques, and algorithms to encode real intelligence in machines to effectively understand and effectively interact with any environments (physical, mental, social, digital, or virtual)
- Real ML is a part of AI that enables computing machinery with the causal power to self-learn from scientific world’s models and experience. It overrules a black box statistic learning ML model, largely neglecting causality, essential to most forms of real learning
- Ultra-DL is a subset of ML that uses causal neural networks to semantically process world’s information using both unmarked and marked training data. The Ultra-DL’s inputs and outputs may be high-dimensional and unstructured data, and inner workings are governed by a comprehensive causal world model.
- Real neural networks are a series of causal algorithms that mimic the operations of a human brain to recognize relationships between real-world data. It recognizes underlying relationships in a set of causal data through a process that mimics the way the human brain operates. [NNs usually trained by a rote learning, by processing samples of a known "input" and "result," forming probability-weighted associations between the two, which are stored within the data structure of the net itself. The training of a neural network is usually conducted by determining the difference between the processed output of the network (often a prediction) and a target output. This difference is the error, or the normally distributed residuals. The network then adjusts its weighted associations according to a learning rule and using this error value. Iterative adjustments will cause the NN to produce output which is increasingly fitting to the target output]. RNN systems learn to perform tasks by considering examples, relying on prior causal models of the world and task-domains. Causal neurons are aggregated into causal layers. Different layers may perform different transformations on their inputs. Causal information signals travel from the first causal layer (the input layer), to the last layer (the output layer), possibly after traversing the layers multiple times. A “neuron” in a causal neural network is a mathematical function that collects and classifies causal information according to specific architectures, as pictured below. The causal network graphs bears a strong resemblance to causal methods and rules reifying statistical techniques and methods such as curve fitting and multiple regression analysis. The goal of multiple causal regression is to model the complex nonlinear relationships among causal variables, as the explanatory (independent) variables and response (dependent) variables.
Between causal neural layers, multiple connection patterns are possible, as symmetrically directed cyclic causal graphs. They can be fully connected, with every neuron in one layer connecting to every neuron in the next layer.
All the partitions of human-like AI, as ML, DL and ANNs, narrow and weak, strong or general, or superhuman, are just redundant.
Ultimately, there is one real and true and genuine AI, the Trans-AI transgressing ML, DL and ANNs, narrow and weak, strong or general, or superhuman AIs.
"Entities should not be multiplied beyond necessity", as states Occam's razor, the principle of parsimony, one of the best and simple problem-solving principles.
The Trans-AI will eat the world
The Real-World, Scientific AI vs. the Unreal, Unscientific AI: Disrupting the Old-Style, Narrow, Weak, Statistic AI/ML/DL