Artificial Intelligence as Religion

BOOK REVIEW: Homo Deus – A Brief History of Tomorrow, by Yuval Noah Harari (2017)

Will the great arc of human history tell us where AI will take us?  If so, Yuval Noah Harari should be a helpful guide.  He is a professor of history, and has received adulation from many quarters, including Barack Obama and Bill Gates, for his earlier sweeping history of mankind, Sapiens.  Harari is an unusual historian in not being abashed by the prospect of speculating on the future, and I think this commendable because, as James Baldwin said “History…does not refer merely, or even principally, to the past. On the contrary, the great force of history comes from the fact that we carry it within us, are unconsciously controlled by it in many ways, and history is literally present in all that we do.”  Homo Deus is beautifully written, very informative, and entertaining.  Its speculations are thoughtful and original.  Harari’s forceful presentation will leave you wondering whether he is prophesying rather than speculating, but the very last few pages make it clear that he sees the possibilities he outlines as just that, possibilities.  And I see them in the spirit of science fiction, to wonder what might come from our technologies based on insight into human nature.  I question many of his ideas, as I will outline below, but applaud the spirit in which they are offered. 

The structure of the book is important.  It is in three parts.  The first part, including a long introduction and Part I: Homo Sapiens Conquers the World, is all history, and is impressive in invoking historical information to form a perspective on how quickly, really in “the past few decades,” humankind has almost overcome the three misfortunes that formerly dominated the entire human agenda: war, famine and plague.  Part II: Homo Sapiens Gives Meaning to the World, elaborates the implications of our history for the future.  And Part III: Homo Sapiens Loses Control, speculates and warns of what things may come. 

The conclusion of Part I is “Sapiens rules the world because only they can weave an intersubjective web of meaning: a web of laws, forces, entities and places that exist purely in their common imagination. “ Intersubjective is the key word here.  It is a concept developed by Harari, consisting of a “third level of reality” beyond objective and subjective entities.  “Intersubjective entities depend on communication among many humans rather than on the beliefs and feelings of individual humans.”  These are stories and myths that are socially reinforced by sharing, and held to be true because of this social reinforcement rather than corresponding to objective truths.  Religious beliefs, for example, are intersubjective realities.  So are money and corporations.  Humans are superior to other animals because they are “the only species on earth capable of cooperating flexibly in large numbers.”.  Bees cooperate in large numbers, too, but “bees did not beat us to the nuclear bomb … because their cooperation lacks flexibility.”  Note that this idea of flexibility implies something like free will, or at least a degree of freedom more than bees have, although later in the book Harari denies the existence of fee will.  (Or appears to.  It’s hard to tell, because so much of the narrative seems to be in the voice of the common accepted truths, which Harari is conveying tongue-in-cheek, rather than Harari’s own conclusions.)   Intersubjective reality is so important to homo sapiens that the “lives of most people have meaning only within the network of stories they tell one another.” The segue to Part II, Humans Give Meaning to the World, is that “In the twenty-first century fiction might thereby become the most potent force on earth. Hence if we want to understand our future, cracking genomes and crunching numbers is hardly enough.  We must also decipher the fictions that give meaning to the world.”  So far, I’m with Harari 100%.  These are important concepts backed up by well-explained history and science.

In Part II, Harari says a religion is “defined by its social function rather than by the existence of deities.  Religion is any all-encompassing story that confers superhuman legitimacy on human laws, norms and values.”  Communism is a religion, as well as Catholicism and Judaism.  So are liberalism, socialism, and Nazism.  There’s a lot of discussion of liberalism as a religion, throughout the remainder of the book.  Humanism is the current reigning religion.  It has replaced gods or God with humans as the originator of its laws, norms and values.  “Modern society believes in humanist dogmas, and uses science not in order to question these dogmas, but rather in order to implement them.” 

Humanism has three variations: communism (or socialism), liberalism, and Nazism.  According to Harari, for “orthodox” humanism, human experience is the ultimately most important reality. “Hence (according to humanism) we ought to give as much freedom as possible to every individual to experience the world, follow his or her inner voice and express his or her inner truth. Whether in politics, economics or art, individual free will should have far more weight than state interests or religious doctrines. … Due to this emphasis on liberty, the orthodox branch of humanism is known as ‘liberal humanism’ or simply as ‘liberalism.’”

I demur on these descriptions of both humanism and liberalism, which form the basis for a lot of what follows in the book.  I’m no philosopher or historian, but my understanding of humanism is that it elevates human reason, as much as or more than individual experience, over religious dogma.  Harari says “It is liberal politics that believes the voter knows best. Liberal art holds that beauty is in the eye of the beholder.  Liberal economics maintains that the customer is always right.  Liberal ethics advises us that if it feels good, we should go ahead and do it.”   My understanding of the reason that the voter knows best is that we are “created equal,” and have equal rights deriving from this fundamental equality, which gives my vote as much importance as anyone else’s.  I find Matthew Stewart’s discussion of how Enlightenment philosophy influenced the liberal political movements of the eighteenth century in Nature’s God ¹ more convincing than Harari’s. Stewart’s contention is that it is belief in human reason vs. religious dogma that is the chief characteristic of these historical liberal movements, and specifically the founding concepts of American independence and law.  Harari says that “feelings,” not reasoning, are paramount for liberals. Curiously, Harari doesn’t discuss equality at all in describing humanism and liberalism in his chapter, The Humanist Revolution, except to mention that the socialist branch of humanism values equality above freedom.  He describes Nazism as “evolutionary humanism,” which celebrates the survival of the fittest but ultimately believes in human experience as paramount in accordance with its humanist origins.  The fundamental weakness of humanism due to its focus on subjective experience makes it vulnerable to displacement by the new data worshipping religion that Harari describes later in the book.

In part III we get to the future.  The central question posed at the beginning of Part III is “How do biotechnology and artificial intelligence threaten humanism?”   Harari says “Liberals value individual liberty so much because they believe that humans have free will.”  Then he goes to great lengths to show that science has established that free will cannot exist, so there is a contradiction with this foundation of humanism. He maintains that the concept of free will was tenable in the 18^th century before fMRI, but not with today’s understanding of brain mechanics.  He claims brain science undermines the concept of intention, and that science, especially the life sciences, “is converging on an all-encompassing dogma, which says that organisms are algorithms and life is data processing” as a major conclusion.  This is presented as “the elephant in the laboratory,” creating a “contradiction between free will and contemporary science.”  He gives a brief mention of the potential for random events at the subatomic level, but says “when random accidents combine with deterministic processes, we get probabilistic outcomes, but this too doesn’t amount to freedom.”  He repeats this assertion that the life sciences in particular have determined that organisms are just algorithms many times.  It is not well supported.  It is a major theme of the final chapter, The Data Religion, but the references in this chapter do not include any from the biological sciences. They are almost all from economists.  The closest thing to life science references are a book by a publicist called Global Brain², and some criticisms of Lysenko. 

Harari introduces the idea that a “Cognitive Revolution transformed the sapiens mind, giving it access to the vast intersubjective realm,” which allowed it to “create gods and corporations, to build cities and empires, to invent writing and money, and eventually to split atoms and reach for the moon.” “As far as we know, this earth-shattering revolution resulted from a few small changes in the Sapiens DNA and a slight rewiring of the Sapiens brain.”  This has led to aspirations that further manipulations of the brain can lead to greater expansions of human consciousness. At the same time, the conflict between science, which purportedly shows that freedom is mythical, and humanism’s fundamental belief in freedom has led to “the great decoupling” between consciousness and intelligence.  The intelligence of algorithms overwhelms the experience of consciousness.  A new religion has arisen, the religion of Dataism, which Harari believes has replaced traditional religions.  “As algorithms push humans out of the job market, wealth and power might become concentrated in the hands of the tiny elite that owns the all-powerful algorithms.”  “Liberal habits such as democratic elections will become obsolete, because Google will be able to represent even my own political opinions better than I can.”

“Techno-humanism faces an impossible dilemma here.  It considers human will to be the most important thing in the universe, hence it pushes humankind to develop technologies that can control and redesign the will.”  But “We can never deal with such technologies as long as we believe that the human will and the human experience are the supreme source of authority and meaning.  Hence a bolder techno-religion seeks to sever the humanist umbilical cord altogether. “ This is Dataism: we, and all organisms, are all just algorithms and the machines we make will do and decide everything for us, making all but the elite redundant.  This is the resolution of the conflict in favor of modern science and at the expense of liberal freedom. Information processing has become superior to mind in our society.  Information itself demands to be free. The elites in control think this is all fine. Authority will shift from humans to algorithms (controlled by the elites). The world will become post-liberal, and “might be an Orwellian police state that constantly monitors and controls not only all our actions, but even what happens inside our bodies and our brains.”  The controlling elite will be augmented, “upgraded.”  This is the Homo Deus of the book’s title.  In the final few pages, Harari holds out some hope: “Maybe we’ll discover that organisms aren’t algorithms after all.” but it seems a thin, unsupported hope. 

I don’t believe there is a scientific consensus that organisms are just algorithms. Harari doesn’t quote any scientists who say that.  Many scientists are religious and believe there are things at work in the universe far beyond what can be described by a simple deterministic algorithm.  They do believe in physical laws that govern all matter and energy including matter and energy in living things, but there are still many mysteries in physical matter and energy, especially at the quantum level that governs the interactions of atoms, molecules and energy, including those in organisms. That’s not the same as reducing an organism to a simple algorithm like a computer program.    

It’s unclear whether Harari himself accepts that science has put the final nail in the concept of free will. He relates details of brain function experiments that show action potentials occurring before subjects declare their intentions, which have been interpreted as proving that intentions are illusory.  He presents these and other information such as fMRI studies showing mechanisms associated with conscious thought as demonstrating that free will is an illusion. Harari concludes that our actions are either pre-determined or random, but never free.  But it is unclear whether he is presenting this as the voice of the current consensus shaped by the Dataist religion, or whether he is buying into it himself.  In any case, I don’t believe that the possibility of free will has been disproven.  The studies cited by Harari and others are never quite clear regarding when an intention is formed vs when it is reported.  Studies of brain function by fMRI and other sensing technologies are very crude.  None of them have yet shown the precise mechanisms by which even straightforward actions such as visual image processing work, much less explain the neuronal mechanism for processing a specific thought.³  Merely showing that some actions, like reflexes, are unintentional doesn’t prove that others can’t be initiated by free will.  It only shows there are constraints on free will, and that some actions are reflexive or driven by unconscious processes, while others may be free.  

And randomness, even if constrained by probabilities, does imply freedom.  Statisticians refer to “degrees of freedom” in random processes. Randomness doesn’t just represent uncertainties caused by lack of knowledge.  Quantum mechanics has demonstrated that there is a fundamental uncertainty in physical interactions.⁴ Things don’t happen until they happen, and there is some freedom in how they will happen until they do.  Scientists and philosophers continue to debate the question of determinism in physical processes.⁵ Determinism and free will are not closed issues in science or philosophy.  

Is it possible that consciousness is closely related to free will?  That the conscious mind is necessarily a mind that freely chooses, rather than just reacting deterministically, and this is the essential quality of consciousness?  Could it be that the probabilities of quantum events override deterministic events in some cognitive brain functions? Roger Penrose speculated on the possibility of quantum state superposition playing a role in brain processes over twenty years ago.⁶  Harari points out that consciousness is not only poorly understood, but that we “don’t have a clue” how it is produced in the brain.  But there has been informed speculation on the subject.  Terrence Deacon’s fascinating book Incomplete Nature – How Mind Emerged from Matter (2012) goes into great detail regarding how organic substances may have assembled themselves into replicating organized cells encapsulating recursive information processes that enhanced their survival.  Could it be that these recursive processes at some point developed the ability to invoke quantum processes, getting a step ahead of determinism and thus stealing a march on more limited organisms?  Could it be that there are degrees of free will, and degrees of consciousness?  We don’t know, and that’s just the point.  We can’t say free will is dead, and it’s valid territory for speculation. 

I think Harari is overstating the case with regard to the forces lining up to worship a data god, with a priesthood that will use algorithms to make everyone redundant except the controlling elite.  There are real dangers of something like this happening, and Harari is right to warn of the trend, which is real.  We need to explore the social power of algorithms and other ways the trend can shake out, good and bad. But the trend is not as simple and clear cut as he makes it seem, and it doesn’t amount to a religion.

What’s most disappointing about Homo Deus is, because of Harari’s focus on proving that Dataism and its worship of soulless algorithms is the great threat today, he doesn’t consider the possibility of consciousness or free will arising in artificial intelligence, as it has in the brain. We really don’t know whether this can happen, and whether, if it does, it might produce a different kind of consciousness than we know. 

Homo Deus is a worthwhile and rewarding book to read, and Harari’s warning of a potential dystopia of a small elite armed with powerful algorithms is valid.  But he seems to have closed off some other possibilities prematurely, and some of them may be just as dangerous, and some may be beautiful and exciting. 

---

[1] Stewart, Matthew, Nature’s God – The Heretical Origins of the American Republic (2014)

[2]Howard Bloom, Global Brain: The Evolution of Mass Mind from the Big Bang to the 21^st Century (2001).  Bloom was trained in biology but spent his career as a publicist for rock bands.

[3] September 2018’s Scientific American carried an article describing a breakthrough in which modified rabies viruses, which defeat the blood-brain barrier by creeping along single neurons and their connecting axons, were used to trace neurons from the retina to their targets in the lateral geniculate nucleus (LGN) of the brain.  The grand result is that scientists have been able to now determine, based on the distribution of targets in the brain, that some sorting of input from left and right eyes seems to occur in the LGN.  This is a great improvement from a 30-year old text on visual processing I recently perused which said that it was clear that the LGN seemed to be the distribution center for retinal signals, but what happened inside it was a mystery.  This is the typical state of the art of our knowledge of information processing in the brain.

[4] Physics Nobel Laureate Steven Weinberg summarizes the current thinking on quantum uncertainty as of 2018 in his recent book “Third Thoughts.”

[5] See Prigogine, Ilya, The End of Uncertainty (1996) for a prominent (Nobel prize winning) scientist’s perspective, and Popper, Karl, The Open Universe, An Argument for Indeterminsm, (1988) for a prominent philosopher’s.

[6] Penrose, Roger, The Emperor’s New Mind (1989), p. 400 and ff.

Real and Artificial lntelligence

BOOK REVIEW: Common Sense, The Turing Test and The Quest for Real AI by Hector Levesque (MIT Press, 2017)

Levesque’s lucid and brief (156 pages) book is an elegant and timely antidote to the overblown hype, snaky language, and grandiloquent assumptions lacing the popular writing on AI.  It’s surprising that much of this overwrought writing is by bona-fide computer science and other prominent professionals.   Levesque uses very little jargon and carefully defines for the general reader what specialized terminology he does introduce.  Going flat out on accessibility, he even makes a point of using almost no math in the entire book, but can’t resist a brief description with simple algebra examples of how math actually works in computing machines, demystifying it a little. 

A major theme is that the current exclusive focus on what Levesque calls adaptive machine learning (AML) in image recognition, self-driving cars, medical diagnoses, and similar applications typically using neural networks, has severe limitations.  It is basically “training on massive amounts of data,” which is a radically different approach than the “good old-fashioned AI” (GOFAI) of the past several decades, which attempted to emulate thinking.  GOFAI sought “common sense,” which quote references an oft-referenced paper by AI pioneer John McCarthy, “Programs with Common Sense,” from 1956, the year of the famous meeting at Dartmouth by some foundational early thinkers about AI (McCarthy, Allen Newell, Herbert Simon, Marvin Minsky et al).  GOFAI was much more concerned with using language, symbols, and knowledge to compute solutions than the current emphasis on machine learning from training on big data sets.

Levesque pays great attention to the question of what we mean by intelligence in ourselves and in machines, distinguishing him from the run of AI savants these days.  Levesque, unlike almost all other AI writers, actually gives a definition of intelligence: “People are behaving intelligently when they are making effective use of what they know to get what they want” (p. 40).  This puts the emphasis on knowledge, stressing the requirement that ability to deduce from knowledge not directly related to a subject under consideration is a critical component of intelligence.  Intelligence handles the unexpected.  It does so through its ability to bring a wide array of knowledge to any problem.  This is “common sense.”  The “knowledge representation hypothesis” is derived from Leibniz and explicated by philosopher Brian Smith (pp 119 – 122).  It’s basic implication for AI is summarized in three lengthy bullets, which I with temerity abbreviate here as:

- An intelligent system must have an extensive knowledge base, stored symbolically.

- The system processes the knowledge base using logical rules to derive new symbolic representations that go beyond what was explicitly represented in the knowledge base.

 - Conclusions derived from the above drive actions.

Levesque shows how AI systems designed based on this representation will be more likely than AML systems to be both reliable and predictable.  To Levesque, reliability and predictability are essential, and the AML approach de-emphasizes them in favor of achieving a preponderance of positive results.

But Levesque doesn’t say that the knowledge representation is all we need, or that AML doesn’t have a place in AI.  He introduces the concept of “The Big Puzzle,” which is that intelligence has many aspects that we don’t completely understand, like different parts of a jigsaw puzzle that we have to solve separately and then bring together.  Parts of the big puzzle include language (symbolic representation), psychology, neuroscience, and evolution, among others.  He notes that one large difficulty in solving the big puzzle is reconstructing a process from its output.  This is inherently difficult, and he demonstrates the point brilliantly with a description of a very simple computer program (introducing neatly some basic algorithmic concepts in the process).  When you see the output, you get his point about how difficult it would be to determine how the program worked, just by examining the output.  Similarly, watching a bird or an airplane fly gives little clue how to design a flying machine. His approach is to take “the design stance,” which he describes by example of the Wright brothers’ approach of understanding in general the processes needed to lift an object moving in air and designing the most practical way to achieve it, rather than learning to fly by imitating birds, which was tried and tried and never worked. He’s saying that neural nets are like trying to fly by imitating birds.  The analogy is limited because neural nets clearly have achieved impressive results, but the results are more like effective data processing rather than what we would call intelligence in a person.   

The Turing test has come to be seen as a valid way to judge whether a computer exhibits human-level intelligence.  Its basis is that intelligence can only be judged by “externally observable behavior,” i.e. its results.  Levesque has an issue with the Turing test.  It requires that the computer be judged on is ability to “fake” human reasoning and common sense, rather than meeting a more objective standard. 

 As a way to overcome the shortcomings of the Turing test, Levesque points out the value of (and has done a lot of research on) Winograd schemas in testing AI systems.  Winograd schemas have simple right and wrong answers. They are statements with a specific but ambiguous pronoun that requires background knowledge or “common sense,” outside the specific facts under consideration, to decide between two specific alternative understandings. Example:

“The trophy would not fit in the brown suitcase because it was too small.”  What was too small?
- the suitcase?
- the trophy?

Levesque questions whether general intelligence AI may even be a worthwhile goal, and notes the trend in actual AI development is for specialized intelligent systems for assistance with specific tasks vs human-level general AI.  He notes that even chess-playing programs are not treated as competition by human players, who today use them for practice and advice in human-to-human competition, which is what they are interested in.  He’s doubtful that anyone will find artificial general intelligence worth paying for.  He also thinks the danger of an autonomous “singularity” taking control is overblown, and shouldn’t be considered on a par with other societal dangers like pollution and overpopulation.  He gives little credence to the idea of superintelligence happening spontaneously or by accident: “Inadvertently producing a superintelligent machine would be like inadvertently putting a man on the moon!”  Autonomy is the real risk, not superintelligence, and autonomy can be carefully controlled.  (Don’t let cars self-drive, for example.  Airline pilots use auto-pilot, they monitor and maintain control over it at all times, never ceding autonomy.)

I’ve been surprised that most of the writing over the past decade on AI is, unlike Levesque, really fuzzy about what we mean by “artificial” intelligence and “general” intelligence, but then goes on to use those terms extensively and as central topics.  If we are going after human-level or “general” intelligence, which these authors commonly assume, or if we’re afraid it may arise spontaneously from our machines, then it seems important to have at least a working definition of intelligence to focus on.  Defining the goal, or the threat, seems an essential first step.  Human-level intelligence is usually considered the model for “general” intelligence, implying that human-level intelligence is something we understand well.  Sometimes there is a discussion of why this is assumed, sometimes involving a discussion of the difficulty of defining intelligence and/or general intelligence, but then concluding that human intelligence is the one we know, so it’s the best reference we have. But how well do we really know it? The idea is then extended to the human brain being the only example and thus model of a general intelligence in nature, eliding a discussion of what it is that makes the human brain uniquely intelligent.  And how?  Exactly how?

It appears the target for general intelligence keeps moving.  When I was a teenager in the 1960’s, it was said that if a machine could be taught to play chess well enough to beat humans, it would be convincing evidence that it had achieved a human level of intelligence.  That goal has been exceeded, but no one is saying that Deep Blue thinks, like a human or otherwise.  Things touted as intelligent behavior at one time are called something else after a machine does them.    

It seems like what is called AI these days by people who are actually making it, rather than writing about it (sometimes the same people in different roles), doesn’t at all aspire to general intelligence.  Instead, its goals are lesser things like greater efficiency in image and speech recogniton, automatic car navigation, unbiased medical diagnostics and similar goals for specific products.  The “intelligence” is usually that the algorithm is deeply heuristic and “learns” or is programmed to try different approaches to the specific problem in order to optimize a solution.  Yet many of the books and commentary on AI (For example, Superintelligence, 2014, by Nick Bostrom, and The Master Algorithm, 2015, by Pedro Domingos, both valuable reading) seem to assume that AI is leading to something akin to “general” human-type intelligence, either by intention or by accident.  Early this year, “Deep Learning: A Critical Appraisal,” (https://arxiv.org/abs/1801.00631) by NYU professor of cognitive science Gary Marcus drew much attention by pointing out that the neural net-based deep learning approach appears to be “approaching a wall,” and “must be supplemented by other techniques if we are to reach artificial general intelligence.”  The analysis is impressive, identifying 10 specific shortcomings, all of which are things that seem worth considering in conceiving of what “general intelligence” might be.  But the ten things don’t add up to anything like a complete description of what general intelligence is.  There were 10 blind men and an elephant…

There seems to be a gap here, that is widening, between what the practitioners are actually pursuing vs a fuzzier but widely assumed goal of general intelligence as not only worthwhile but the ultimate goal.  One could conclude from this that the goal (or threat) of artificial “general” intelligence at “human-level” or above is proving to be a chimera (or a false threat).  But it could also be true that more clarity around what we mean by “general” and “human-level” intelligence in a machine would go a long way toward helping us see the real value of deliberately pursuing a more general artificial intelligence, and how to achieve it, as well as the real danger of its threat.