Reifying AI, ML, DL and ANNs: Causal Machine Intelligence and Learning (CMIL)
The world is flooded in uncertainty, and only very few things are truly predictable. Despite its apparent chaos, the world is deeply ordered by fundamental principles, rules and laws helping humanity to discover ways to survive and thrive. And a critical role has always belonged to the human minds, brains, intellect and intelligence. The more intelligence, the more intelligent, reasonable and rational life. Due to its knowledgeable and innovative power, the human intelligence is now creating its best ever partners and companions, powerful machine/computer intelligence, as artificially intelligent technological systems, all studied and developed by AI Science and Engineering (AISE).
It is forming as a transdisciplinary science and engineering of?[Technology Systems, Platforms, Machines, Computers, Computer-Controlled Robots, or Software Powerful to Interact Effectively, Sustainably and Intelligently with the World]
The key for any powerful intelligence, human, machine, or alien intelligence, is true, comprehensive, coherent and consistent mapping, modeling, simulating or "understanding" of reality and causality, causation or cause and effect.
Causal Structure Discovery Algorithms: ML Inferring Causal Structure/Model From Data?
Everything is learned bottom-up from data, observations and experiences or everything is learned top-down from rules, templates, basic facts or ground truths.
The first paradigm is what about today's empirical science and its poor imitations, machine learning models, hunting after individual observations or experiences, which are merely instantiations of classes and rules or some ontological templates and scientific abstractions, be it naive physics facts or general rules of the commonsense world.
This is what grounds a statistic-data-base AI/ML/DL unable to learn true facts top-down, but only learning bottom-up, from instantiations to general templates.
The second paradigm is what qualified as a non-science or speculative theoretical science. In fact, any real comprehensive learning of basic facts is impossible without generalizations, and high-level concepts and symbolic representations.
Empiric scientific research is predominantly focused on statistical correlations and functional relations instead of real, causal relationships, with causal discovery, model, reasoning, inference.
The key for any powerful intelligence, human, machine, or alien intelligence is true mapping, modeling, simulating or "understanding" of reality and causality, causation or cause and effect, to effectively and sustainably interact with the world.
In reality, hypothesis-driven research should assume a causal structure, a set of causal relationships among inputs and outcomes, and researchers estimate the effect size of these relationships (e.g. causal inference), as structural equation models (SEM). In such research, drawing a causal conclusion is valid, because prior knowledge ascertains that the relationships are indeed causal (rather than merely associative or correlative). And when there is no knowledge of the causality, the causal structure itself needs to be discovered from data through a process known as causal structure discovery.?
Causal structure is the set of causal relationships among a set of variables, and causal structure discovery (CSD) is about learning the causal structure from a stream of real-world data, observational, experimental or computational.
CSD is about identifying causal relationships from large quantities of data through computational methods.
Instead of traditional association-based ML computational methods to discover data patterns, (i) CSD methods can discover the unknown causal relationships from observational data and (ii) to offer guidance to accurately discover unknown causal relationships.?
?As the “gold standard” graph serves a fully connected undirected causal graph, with all edges connecting conditionally all variables, dependent or independent variables.
To have Real AI as Causal Machine Intelligence and Learning, Causality or Causation should be modelled as an interrelationship between input variables X and output variables Y such that changes in X lead to changes in Y, and vice versa.
The universal abstraction or complex nonlinear causality is supported by a lot of facts and evidences and rules, as in every domains or fields of knowledge and practice:
complex behaviors,
interactions,
interfaces,
circular causality,
reciprocity, mutuality,
nonlinear phenomena in physics, biology, and social sciences,
chemical reactions,
cybernetics, self-regulating mechanisms, feedback loops between processes or inputs and outputs, qualitative or quantitative, positive/negative?with?self-reinforcing/self-correcting,?reinforcing/balancing,?discrepancy-enhancing/discrepancy-reducing?or?regenerative/degenerative,
biofeedback,
social feedback,
information feedback,
nonlinear narrative in literature and cinematography,
the special backpropagation algorithms for training neural networks, reinforcement learning,
and Bayes' theorem (Bayes' law?or?Bayes' rule; Bayes–Price theorem)
True causality is fundamentally symmetric, i.e., causes lead to effects and the other way around.
Real and true and genuine CAUSALITY is the INTERRELATIONSHIP of cause and effect, causality interrelates or is interrelated; interacts or is interacted.
CAUSALITY IS THE INTERRACTION OF CAUSE AND EFFECT, when an action is influenced by other actions as reactions.
CAUSATION IS AN INTERACTION, when?two or more entities have an effect upon one another.
Causative interaction or Interactive Causation or Interaction is the key mechanism of the world and the key construct of all sciences and technologies, physics, chemistry, biology, cognitive science, sociology, economy, or computer science and engineering.
In physics, there are four known fundamental interactions: the?electromagnetic,?strong,?weak?and?gravitational?interactions.? And a theory of everything is where all its fundamental interactions are unified. It is hypothesized that the combinations of many simple interactions can lead to emergent?phenomena, as complex systems or integrative levels of organization, the strata of layers. It structures reality, arranging all entities, structures, and processes in the universe, or in a certain?domain, into a hierarchy, like matter, life, mind, society and machine superintelligence,
In all, it implies all kinds of interrelationships, as reciprocal or mutual relations:?connection, connect, interaction, interconnection, interdependence, interdependencies, interlink, interplay, inter-dependencies, correlation, linkage, inter-relation, interrelatedness, or pattern and order.
True causality is an integrated relation and interactive relationship. It is nonlinear by the very nature, where
Cause Causes Effect if and only if Effect Causes Cause, or X causes Y iff Y causes X
That means: a variable, X, can be said to cause another variable Y, if when all confounders are adjusted,?an intervention in X results in a change in Y, and an intervention in Y does necessarily result in a change in X. This is in line with correlations, which are inherently symmetric, i.e., if X correlates with Y, Y correlates with X, as if X causes Y, Y MUST cause X.
Bayesian and causal networks are completely identical. And the difference lies in their misinterpretations. Consider the simple example in the figures below.
Example network that can be interpreted as both Bayesian and causal. Fire and smoke example adopted from Pearl [The Book of Why: The New Science of Cause and Effect].
Here we have a network with 2 nodes (fire icon and smoke icon) and 1 edge (arrow pointing from fire to smoke). This network can be both a Bayesian or causal network.
The key distinction, however, is when interpreting this network. For a?Bayesian?network, we view the?nodes as variables?and the?arrow as a conditional probability, namely the probability of smoke given information about fire. When interpreting this as a?causal?network, we still view?nodes as variables, however the?arrow indicates a causal connection. In this case both interpretations are valid.
Now, if we were to revert the edge direction, the causal network interpretation would be still valid, regardless that smoke does not?cause?fire.
Fire is the rapid oxidation of a material (the fuel) in the exothermic chemical process of combustion, or burning, releasing heat, light, and various reaction products, including gaseous by-products, as smoke. It is commonly an unwanted by-product of fires (including stoves, candles, internal combustion engines, oil lamps, and fireplaces). Smoke is not included in the fire tetrahedron, oxygen, heat, fuel and chemical reaction. Take any of these four things away, and you will not have a fire or the fire will be extinguished.
Fire causes heat and light as much as light or heat could cause fire. Again, if the Fire is a rapid oxidation of a material, their must be an inverse process of deoxidation/reduction which occurs when there is a gain of electrons or the oxidation state of an atom, molecule, or ion decreases. ?Or, as to the inverse causality law, if one chemical species is undergoing oxidation then another species undergoes reduction.
So, X causes Y if an intervention in X results in a change in Y, and an intervention in Y does necessarily result in a change in X.
That all means that we have one of biggest human misconceptions, cognitive biases or informal fallacies, misguiding human decision-making and practice for all the human history.
Again, a statistical input-output dependence or asymmetric functional relationships of dependent and independent variables as used in mathematical modeling, statistical modeling and experimental sciences should be reviewed.
Depending on the context, an independent or exogenous variable is called a "predictor variable", regressor, covariate, "control variable" (econometrics), "manipulated variable", "explanatory variable", exposure variable, "risk factor" (medical statistics), "feature" (in machine learning and pattern recognition) or "input variable".
A dependent endogenous variable is called a "response variable", "regressand", "criterion", "predicted variable", "measured variable", "explained variable", "experimental variable", "responding variable", "outcome variable", "output variable", "target" or "label".
Causal relationships?between variables in real-world settings are much more complex and may consist of direct and indirect effects, that go directly from one variable to another or when the relationship between two variables is mediated by one or more variables.
Establishing?delineated causal relationships?is a key method of empirical research in natural and social sciences, manipulating the independent variable in order to determine the impact on a dependent variable in the control settings (laboratories).In fact, specific causal links from one variable,?X, to another,?Y, cannot usually be assessed from the observed association between the two variables. The reason is that a big part of the observed association between two variables may arise by?reverse causation?(the effect of?Y?on?X) or by the confounding effect of a third variable,?X, on?Z?and?Y.
"Consider, for example, a central question in education research: “Does class size affect test scores of primary school students? If so, by how much?” A researcher may be tempted to address this question by comparing test scores between primary school students in large and small classes. Small classes, however, may prevail in wealthy districts, which may have, on average, higher endowments of other educational inputs (highly qualified teachers, more computers per student, etc.) If other educational inputs have a positive effect on test scores, the researcher may observe a positive association between small classes and higher test scores, even if small classes do not have any direct effect on students' scores. As a result, observed association between class size and average test scores should not be interpreted as evidence of effectiveness of small classes improving students' scores.
This gives the rationale for the often-invoked mantra “association does not imply causation.” Unfortunately, the mantra does not say a word about?what?implies causation. Moreover, the exact meaning of causation needs to be established explicitly before trying to learn about it".
That's biggest, badly entrenched misconception; for "causation implies correlation", and vice versa, association implies causation.
Modeling Real Causality for Real AI
Causality is represented mathematically via?Structural Causal Models (SCMs), with two key elements, a graph and a set of equations.
The golden standard causal graph, C = <E, R>, is a?Bidirected Cyclic Multi-Graph or Causal Loop-Graph Network (BCG/CGN), where?entity-vertices E (circles, nodes, points) in a causal BCG represent variables and edges R (arrows, links, ties, arcs, lines) represent causation, direct or inverse.?
Again, the BCG/CGN embraces all Cause Analysis Tools, from the fishbone and scatter diagrams, Pareto chart to as interrelationship diagraphs, relations diagrams or digraphs, network diagrams, or data matrix diagrams, defined as a?new management planning tool?that depicts the relationship among factors in a complex situation showing cause-and-effect relationships.
Fishbone Diagram Example
The Causal Network Graphs as a data matrix diagram is used for analyzing and displaying the relationship between data sets, showing the relationship between two, three, or four groups of information and giving information about the relationship, such as its strength, of the roles played by various individuals or measurements. Six differently shaped matrices are possible: L, T, Y, X, C, and roof-shaped (X<>?X when also X<>?Y in L or T), depending on how many groups must be compared.
It is applied in the root cause analysis for a problem or situation to understand links between ideas or cause-and-effect relationships, how different aspects of the problem are connected; to see relationships between the problem and its possible causes that can be further analyzed.
The Causal Network Graphs are strongly connected?containing a directed path from?x?to?y?(and from?y?to?x) for every pair of vertices?(x,?y), while having circuits or loops, that is, arcs that directly connect nodes with themselves, and multiple arrows with the same source and target nodes, thus covering all possible directed graphs as a Directed Acyclic Graphs (DAG) or weighted directed graphs/networks .
The set of equations is a?Structural Equation Model (SEM), but showing the causal interconnections?and?the details of the relationship. SEMs?represent all possible interrelationships between or among variables. These equations are symmetric meaning equality works in two directions. This has the implication that SEMs can be inverted to derive alternative SEM equations.
If the Real AI models the relationship between a disease and the symptoms it produces, Y = mX + b, it also accounts for an inverted relationships between symptoms and diseases, X = (Y - b)/m. Or, if diseases cause symptoms, then we have to interpret the second equation as the symptoms cause diseases!
In the case of modeling causation as a linear relationship between two variables X and Y such that changes in X lead to changes in Y, we have deeply defective causal models, reasoning, inference and causal discovery algorithms. [Pearl, J.?Causality: Models, Reasoning, and Inference. Vol. 64 (Cambridge University Press, 2000)].
In other words, The Book of Why: The New Science of Cause and Effect?by Judea Pearl, basing on a deficient asymmetric causality and DAG/SEM, causal inference and causal discovery, is resulting in scientific malpractice and flawed methodology for statistics, epidemiology, social sciences, eco\nomics, political sciences, computer science, AI and ML.
For today's AI & ML & DL & ANNs to be upgraded to the top level of Causal Machine Intelligence and Learning (CMIL) Systems, the Neural Networks are be reified as Causal Probabilistic Networks (All-Directed Cyclic Causal Golden Standard Graphs).
Causality as the topmost ontological facts
Real Causality could be modelled as Tautologies possessing as the key features Antisymmetry, Transitivity, Reflexivity, and Symmetry or Equivalence marked by Transitivity, Reflexivity, and Symmetry, added with Antisymmetry, satisfying these axioms:
What's wrong with an acausal AI?
Today's AI/ML/DL is revolving around statistics algorithms and correlations, being deficient of causality, causal discovery, causal models and causal inference, aiming to?answer questions involving cause and effect. Causal AI?aims at answering causal questions?as opposed to the statistical or acausal AI.
Often in the real world, however, we may not be sure about which variables cause each other. Causal AI is to?infer causal structure from data. Given a dataset,?its causal discovery algorithms derive?a causal model that describes it.
Now, how messy and confusing are today's ML/AI Models, one could see from the know-how requests to build them:
How to define an ML problem?
How to select a dataset?
How to perform data preparation?
How to perform feature engineering?
How to select an ML algorithm?
How to choose a performance metric?
How to train an ML model?
How to present ML results?
How to design and deploy an ML solution?
How to debug ML models?
There are other topics you will encounter:
Parametric vs Nonparametric Algorithms
Supervised vs Unsupervised Algorithms
The Bias-Variance Trade-Off
Traps of Statistic ML, Spurious correlations, Simpson’s Paradox
How to Diagnose/Fix Overfitting and Underfitting?
How to create a data pipeline?
How to deal with small datasets?
领英推荐
How to deal with imbalanced datasets?
Anomaly Detection
Large scale pattern recognition on suitably collected independent and identically distributed (i.i.d.) data
AutoML
And this you should know before learning AI/ML:
Calculus
Discrete Mathematics
Linear Algebra
Probability and Statistics
Statistical Programming
A central problem for AI and causality is, thus, causal representation learning, the discovery of high-level causal variables from low-level observations.?Learning Non-Linear Causal Relations at Scale, Causal Variables, Inductive Biases, and Causally Correct Models of the World and the Agent are essential for robust and versatile model-based reinforcement learning.
It is a shaky way to simulate/replicate/imitate human mind/intelligence/brains with all the mental capacities, as far as most mentality is a back box and unconscious phenomena.
To build Real AI, or Causal Machine Intelligence and Learning, we should follow reality and causality with its truest representations by transdisciplinary science and engineering, as integrating:
Philosophy, Ontology, Epistemology, Logic, Ethics
Scientific computing or computational science
Computer Science
Cognitive Science, Linguistics, Psychology
Mathematics, Statistics
Natural Sciences
Engineering, Automation, Robotics…
Intelligent technology systems are then technological systems (machines, computers, robots, applications) applying causal AI technology [as the sum of theories, tools, and techniques, models and algorithms, skills, methods, and processes] by taking real-world inputs (data, signals, quantities, information), changing them according to the system's CDA (causal discovery algorithms), to produce a meaningful outcome (classifications, predictions, recommendations, knowledge, products, services, or processes and actions). ?
Causality of Machine Intelligence and Learning
AI/Data Science/ML/DL/Robotics are a 3-body problem, causally interrelating Data, Mind (AI) and Reality.
All the mess-up is coming from splitting one great idea into two opposing types of theories of causation:
The Hume’s theory holds that causation is entirely constituted by facts about empirical regularities among observable variables; there is no underlying causal nature, causal power, or causal necessity.
The causal realist takes causal mechanisms and causal powers as fundamental, and holds that the big task of science and technology is to arrive at empirically justified theories and hypotheses as well as machines and mechanisms about those causal mechanisms.
Briefly, things, entities and events to be such have causal capacities: in virtue of the effective properties they possess, the things in the world have the power to bring about other events or states.
Now causal relations cannot be directly inferred from correlations, data patterns or associations among variables.
In all, there are various assertions about the statement, X caused Y:
Conditional probability sentences, as in?P(y|x) =?p, stating that: The probability of event?Y=y, given that we observed event?X?=?x?is equal to?p. In large systems, such sentences can be computed through Bayesian networks or any number of machine learning techniques.
WHAT IS WRONG WITH ML & DL & AL SYSTEMS
Such machine learning systems lack adaptability, or robustness, the critical ability to recognize or react to new circumstances they have not been specifically programmed or trained for. Here come intensive theoretical and experimental efforts toward general "transfer learning," "domain adaptation," and "lifelong learning" to overcome this obstacle.
Another obstacle is transparency or "explainability," or that "machine learning models remain mostly black boxes" unable to explain the reasons behind their predictions or recommendations, thus eroding users' trust and impeding diagnosis and repair.
In response, the Defense Advanced Research Projects Agency (DARPA), is overseeing the suitably named?Explainable Artificial Intelligence?program.
A third obstacle concerns the lack of understanding of cause-effect connections, the hallmark of human and machine minds, a necessary and sufficient ingredient for achieving human-level intelligence or beyond.
Causal reasoning/inference/forecasting presupposes the causal mechanism. The researcher ought to attempt to identify the deep causal mechanism relating the variables of study testing or confirming a causal hypothesis: apply methods of similarity and difference as a test for necessary and sufficient conditions, examine conditional probabilities, examine correlations and regressions among the data or the variables of interest, etc.
Causal realism posits that empirical evidence must be advanced to test/confirm/assess the credibility of the causal mechanism that is postulated between cause and effect.
There is a three-level hierarchy that restricts and governs inferences in causal reasoning:
1. Association, 2. Intervention, and 3. Counter-factual.
Association invokes purely statistical relationships, defined by the naked data.
For instance, observing a customer who buys toothpaste makes it more likely that this customer will also buy floss; such associations can be inferred directly from the observed data using standard conditional probabilities and conditional expectation. Questions at this layer, because they require no causal information, are placed at the bottom level in the hierarchy. Answering them is the hallmark of current machine learning methods.
The second level, Intervention, ranks higher than Association because it involves not just seeing what is but changing what we see. A typical question at this level would be: What will happen if we double the price? Such a question cannot be answered from sales data alone, as it involves a change in customers' choices in reaction to the new pricing.
Finally, the top level invokes Counterfactuals, a mode of reasoning that goes back to the philosophers David Hume and John Stuart Mill and that has been given computer-friendly semantics in the past two decades. A typical question in the counterfactual category is: "What if I had acted differently?" thus necessitating retrospective reasoning. Counterfactuals at the top of the hierarchy because they subsume interventional and associational questions. If we have a model that can answer counterfactual queries, we can also answer questions about interventions and observations.
How To Create Real AI or Causal Intelligence
Focus on three key things: Reality (Ontology & Science), Data (Computing & Mathematics & Statistics), and Digital Machine Intelligence (Computational science and computer science)
Keep in the mind some simple basic truths.
AI is a complex of hardware and software to effectively and sustainably interact with the world by intelligent modeling and simulating reality in all its complexity.
Real AI is not about replicating/imitating/simulating human intelligence/mind/brains/cognition/…
Real AI is not ML, not DL, not ANNs, not Predictive Analytics, not Programming Languages, not Applied Mathematics, not Computer Science, not Cognitive Science, but transcends all of them as special tools, theories, techniques, models, algorithms or technologies..
The whole idea of Artificial Human Intelligence, be it AI, ML, DL, AGI, or ASI, is the summit of human prejudices to anthropomorphize everything around us, including machine/computer intelligence.
Follow reality and causality with its best representations, real ontology and transdisciplinary science and engineering, as scientific computing or computational science and computer science.
Study the next gen AI, Causal Machine Intelligence and Learning Systems, with Artificial Neural Networks to be reified as Causal Probabilistic Networks (All-Directed Cyclic Graphs).
Conclusion
Causal AI, Real AI, Causal Machine Intelligence and Learning, True Computer Intelligence, Causal Deep learning and Artificial Neural Networks, Meta-AI, Trans-AI, Metaverse, one needs to learn it all not to become a relict.?
Resources
SUPPLEMENT 1: AI and BI
The whole idea of Artificial Human Intelligence, be it AI, ML, DL, AGI, or ASI, to mimic/simulate/replicate biological intelligence is?the summit of human prejudices to anthropomorphize everything around us, including machine/computer intelligence.
Our brains fall under the special class of biological intelligence, which has been evolving over a billion years of trials and errors, in various forms and kinds.
There is a smart piece comparing BI with AI. Here are some key points:
1. Biological intelligence engages all the conscious and unconscious knowledge of a human being. That immense field stretches from genetics to culture to society and psychology. Much of it is hardly understood. Your mother’s arm that holds you in an embrace, the lover’s hand that gently touches your cheek, and the little gestures that tell you’re loved will prove hard work for robots.
2. Biological intelligence is connected to everything inside you—every information system you use. You have an immune system, a cardiovascular system, a hormonal system, a muscular system, dozens of interconnected systems. Unlike most robots, the body doesn’t do one thing at a time. It coordinates all the different information systems at the same moment. Can you presently conceptualize the systems that will have a robot laugh, cry, sing and dance, all while gauging the audience and telling a joke? Comedians do that. The size of your body’s information systems dwarf the complexity of the entire Internet. Our medical attempts to make ourselves not ill are generally far less impressive than the actions biological intelligence engages every moment to keep us healthy.
3. Biological intelligence has different goals than artificial intelligence. What is it for? In our case, the survival of our species. Biological intelligence is built to keep humanity going. Normally that includes you, and me, and everyone we know—but not always. Having genes for sickle cell may help a population survive malaria, but can really hurt you in places without it. Diabetes genes may keep us stay alive during famines, but otherwise can really mess up your life. Biological intelligence wants the?species?to survive—not just us.
4. Unlike the artificial intelligence you experience in glitching software, biological intelligence has survived almost everything thrown at it. It survived the asteroids that wiped out the dinosaurs. It survived volcanoes that scorched and burnt the earth for millions of years. It survived plagues and pestilence. Most of the species on the planet are gone. Billions of species have disappeared. We’re still here. Why? Because biological intelligence built us to survive.
5. So how does biological intelligence work? Here’s the real trick: biological intelligence is built on contingency and chance. Stephen Jay Gould and others pointed that out long ago. Not only do we have genetic information systems that survived asteroids. We’re built to survive comets, earthquakes, cataclysms and catastrophes that we’ve never seen—and that may never happen. Biological intelligence provides us genes and physiology that's built to survive stresses that do not yet exist and may never exist. Chance rules the world, and we are built to survive all that chance can throw at us. Think of new illnesses, like AIDS. When AIDS first hit it was terrifying. Yet many of us had inbuilt systems to keep it off, even before one effective drug was produced.
6. Unlike artificial intelligence, biological intelligence does not operate just within us, but over a huge ecosystem. That ecosystem is you. There are at least 40 trillion bacteria in your gut. They not only digest food, but now appear to change your mood, your ability to fight off infections, how cancer drugs work. There’s at least 10 times more non-human cells in your body than human ones.
Biological intelligence rules them all.
How does biological intelligence differ from artificial intelligence?
SUPPLEMENT 2
The Book of Why: The New Science of Cause and Effect?Hardcover – May 15, 2018
The book has a false assumption, "Correlation is not causation", badly misconceiving the great idea of causality, causation or cause and effect, corrupting many good minds as one could read below. ?
Editorial Reviews
Review
One of?Science Friday's?"Best Science Books of 2018"
"Illuminating... The Professor Pearl who emerges from the pages of?The Book of Why?brims with the joy of discovery and pride in his students and colleagues... [it] not only delivers a valuable lesson on the history of ideas but provides the conceptual tools needed to judge just what big data can and cannot deliver."―New York Times
"Cause and effect is one of the most heavily debated, difficult-to-prove things in science and medicine. This book really gets you thinking about cause and effect as it applies to issues of our time, such as: How come cigarettes were around for years and we never showed they were causing cancer or heart disease? The authors goes through these cases like an interrogation, and it's just extraordinary."―Science Friday
"Seriously, everyone should read?The Book of Why."―Jeff Witmer,?American Mathematical Monthly
"'Correlation is not causation.' That scientific refrain has had social consequences...Judea Pearl proposes a radical mathematical solution...now bearing fruit in biology, medicine, social science and AI."―Nature
"Lively and accessible...Pearl was one of the visionary leaders of the causal revolution, and?The Book of Why?is his crowning achievement."―Jewish Journal
"Anyone interested in probing connections between cause and effect, and their relevance for the future of AI, will find this a fascinating and provocative book. Highly recommended."―CHOICE
"Judea Pearl is on a mission to change the way we interpret data. An eminent professor of computer science, Pearl has documented his research and opinions in scholarly books and papers. ... With the release of this historically grounded and thought-provoking book, Pearl leaps from the ivory tower into the real world...Pearl has given us an elegant, powerful, controversial theory of causality."―American Mathematical Society
"Have you ever wondered about the puzzles of correlation and causation? This wonderful book has illuminating answers and it is fun to read."―Daniel Kahneman, winner of the Nobel Memorial Prize in Economic Sciences and author of Thinking, Fast and Slow
"Pearl's accomplishments over the last 30 years have provided the theoretical basis for progress in artificial intelligence... and they have redefined the term 'thinking machine.'"―Vint Cerf, Chief Internet Evangelist, Google, Inc.
"Judea Pearl has been the heart and soul of a revolution in artificial intelligence and in computer science more broadly."―Eric Horvitz, Technical Fellow and Director, Microsoft Research Labs
"If causation is not correlation, then what is it? Thanks to Judea Pearl's epoch-making research, we now have a precise answer to this question. If you want to understand how the world works, this engrossing and delightful book is the place to start."―Pedro Domingos, professor of computer science, University of Washington, and author of The Master Algorithm
"The Book of Why ...?questions and redefines the building blocks of our AI systems"―theverge.com
Only a few have noticed its real value: I hate to tell you this, but science, at least real science, has linked cause and effect.?
What the entire farce is doing is called obfuscation. If you are so confused by the math, technical jargon, sciency graphs and tables and data and figures, then you just feel dumb and agree with whatever idiotic conclusion the author invents. Look how cutting taxes and increasing federal spending stimulates the economy with all my sciency charts and formulas! It's an entire scam industry and this author is just like another grifter.