What Minds What We Matter?

4,733 words

Part 2

Jacques Luzi
Ce que l’Intelligence artificielle ne peut pas faire
Èditions La Lenteur, 2024

While scientists and economists are often prepared and willing to debate the “dangers” of specific potential AI developments, they are less disposed to debate the ultimate purpose of AI or whether AI has an ultimate purpose at all. AI is a reaction to challenges, but progresses as blindly as biological evolution itself. By solving practical challenges at an ever accelerating rate and with ever increasing power, AI may take us we don’t know where and we don’t know why. The response to the question “where is AI heading?” is “let’s wait and see”. But is that the answer a patient on an operating table will be happy to receive if he asks  “what’s going to happen to me doc?” Or, as an old joke goes, the patient on the operating table will hardly be reassured if the surgeon tells him in confidence, “don’t worry, it’s my first time too.”

There certainly have been prominent AI advocates who acknowledge specific dangers of AI.  Luzi mentions several such self-proclaimed “concerned” advocates, for example, the scientists Daniel Dennett and Paul Macready.  Pioneers in the field of AI, they both claimed to be deeply concerned about the dangers caused by technological progress and the negative effect of inventions and interventions on the natural environment. Their solutions were more and better technology.  Daniel Dennett, who argued strongly that the brain was no more than a computer, acknowledged that AI could be, in his words, a “danger to civilization”. The advocates of the most modern scientific innovations often claim that the technology they advocate is a response to a challenge: wind turbines (which Paul Macready helped to develop) and solar energy are promoted as a response to the challenge of pollution (and more recently “to meet the challenge of global warming”) caused by fossil fuel consumption, and so on. Luzi regards such concerns as hypocritical and fraudulent. He is highly suspicious of claims made by “sustainable growth” enthusiasts and modern environmentalists who announce that technology will solve ecological problems. Luzi  notes that the case for AI as a solution constitutes an appeal to the need for a solution to a problem created by scientific intervention in the first place. The burglar is proffering advice on solving security issues. For Luzi, a demand for “sustainable growth” is a pretext to justify subventions and support for environmentally deleterious, because more intensively artificial, “solutions”.

One might argue that the emergence of AI is similar to the origin of species by natural selection, but Luzi believes that technology is an acceleration in an unnatural manner, unlike a process of natural selection, where hundreds of thousands of years would pass, leading gradually to adequate adaption by fauna and flora (perhaps it is no coincidence that his book has been published by a publishing house called La Lenteur, Slowness!). Of course, natural catastrophes can make a near clean sweep of many species and make way for a new evolutionary renaissance of diversity, as described years ago by Peter Ward in The End of Evolution, but that begs precisely the question: do human beings want to pave the way to their own extinction to make way for other species more diverse, more benign and more advanced? If so, they would be the first species in the history of Earth to do so.  As Luzi reminds the reader, as if he needs reminding, that the nuclear bomb, that supreme scientific achievement, can destroy Homo sapiens entirely, along with all advanced life forms.  AI may one day enable decisions about total destruction to be taken by unemotional inorganic intelligence.  Luzi is not seeking reassurances about some aspects of AI. He rejects it in all its permutations. An extreme position that may be, but it behooves advocates of AI to show why AI is not the enemy of mankind and why it can’t become so, or if there is a threat, what can be done to counter that threat.

Any consideration of the desirability or otherwise of Artificial Intelligence should be accompanied by an exact understanding of what one means by the term. Artificial Intelligence is usually described in respect of three kinds:- Narrow or Weak AI (ANI), General AI (AGI), and Artificial Superintelligence (ASI).  The first kind (narrow or weak are surely misnomers) includes all forms which execute tasks fed into it, where the task consists of solving algorithmic problems, but these can include complex tasks such as driving a car or the services offered by AI tools such as Grammarly or Chat GBT. This kind of intelligence is based on mathematical logic. If x is presented as a problem, y should be taken as a solution. If x+f is the problem, then y  and so on, becoming ever more complex, but working on the same principle of neurological connections as those made by the human brain. At this stage AI is entirely dependent on the information with  its creators provide it with, and is incapable of independent thought.

AGI is regarded as the second level of AI evolution (and evolution it is) in which machines will be able to interpret human emotions and not just reason and act upon those emotions but feel and express emotions themselves, like Hal in Stanley Kubrick’s 2001 (“I am afraid. I am afraid.”). They might be described as independent thinking and independent thought and would include experiencing the urge to survive and acting accordingly. Some would say that the urge to survive defines life as opposed to non-life.

ASI is ill-defined, being still out of reach. It is understood to mean intelligence which could react and make decisions faster and more efficiently (that word again!) than our species on the basis of best interests of either its creator or even itself and through an evolutionary process would possibly even plausibly supersede Homo sapiens entirely once its cognitive complexity enabled it to become fully self-aware. In an ASI dominated world, it is not inconceivable that Homo sapiens might be superseded by a more sophisticated technical inorganic AI driven entity.

It is a characteristic of debate concerning AI and AI technology that there are many more or less similar definitions of AI, but in their haste to define and usually be “very excited” about AI, technicians and scientists spend less time seeking a consensus on the meaning of the two separate words contained in AI, namely “artificial” and “intelligence”.

Dictionaries usually indicate that the word “artificial” refers to that which does not arise spontaneously in the course of nature but is contrived by man. However, in this reviewer’s opinion confining the meaning to refer to human activity alone is too restrictive.  According to this definition of artificiality, when a human uses a stone to crack an egg, the stone is used “artificially” but when an Egyptian Vulture uses a stone to crack an egg (which that bird characteristically does) no “artificiality” takes place. Similarly: when Australopithicus used a stone tool to cut meat, was the stone tool “natural” on the grounds that Australopithicus was not (yet) human, whereas when a Neantherthal used a stone to cut meat, was the stone tool  “artificial” on the grounds that a Neanderthal was a race of Homo sapiens?  Why should a bird’s nest be less “artificial” than a turf hut or an igloo? Perhaps there is more of the natural in the artificial or more of the artificial in the natural than either extreme advocates or extreme opponents of AI wish to concede.

For the purposes of practical use, the concept of being artificial is best understood to refer to something which intervenes in nature, trying to replace or improve upon the pre-existing model or mechanism. The “natural” can be understood as being that which arises in the natural course of events independently of life forms, while what is “artificial” is what is used by a life form to put something to use as a tool or create a new composition, one not found in nature, for an end which is extraneous to the nature of the object so used. Seen in this light, artificial objects are not dependent on Homo sapiens, yet the use and above all creation of them is very much characteristic of the species. Artificial objects and processes are not then by definition entirely alien to nature, but Luzi thinks they are.

It is relatively easy to identify for practical purposes what is artificial as opposed to what is natural, although gray areas exist.  On the other hand, intelligence is a question of degree. Not only are some people generally and obviously more intelligent than others, there are not only different levels of intelligence, there are also different kinds. For example, there is reflective intelligence and intelligence related to rapidity of deduction, and so on. However,  the dictionary definition of intelligence as the ability to acquire, understand and use knowledge seems to this reviewer less problematic than common definitions of the meaning of artificial. Nevertheless, there are problems here too. Are honey bees, which learn from other honey bees by a system of signaling and dancing, being “intelligent” when they absorb information? By the common definition given to the word they must be, for they are acquiring knowledge and putting the knowledge they acquire to practical use. If we talk about human intelligence, perhaps we must add to “the ability to acquire and use knowledge” the words  “and the ability to assess and evaluate such knowledge with some degree of self-awareness.”  It is self-awareness which presumably would be the line to be crossed for AI to present itself as ASI. For the non-believer, self-awareness arguably approximates to what religious folk call the soul. If that is so, Luzi presumably believes that because it is soulless, AI can never become self-aware. But in that case would it ever be powerful enough to supplant Homo sapiens?

Artificiality and intelligence are inter-connected and in a sense that is independent of any notion of artificial intelligence. What they both have in common is an adaption to nature to reach a perceived end. Wolves and many other animals use pack hunting technique to exhaust prey. One wolf chases prey until the wolf becomes tired and another member of the pack takes over. The isolated prey only has x energy. The hunting wolves have x energy times the number hunting in the pack. This is “intelligent” behavior, but by extrapolation arguably “artificial” too, in the sense that the wolves are going beyond the individual hunting capacity of each one them, to create a hunting pack which is an entity more efficient, one is tempted to say even more intelligent, than the sum total of individual wolves.

Artificiality and intelligence are present in nature and it might be argued that the very paradox of life is that they are so. Artificial intelligence can therefore be defined as an intelligence applied outside the inherent or natural intelligence of the human species (so far only humans are known to have created AI, but who knows what lies in the vastness of space?) to respond and act upon the environment in what its creators perceive to be in the interests of themselves or their species, or, as some believe (and Luzi is one of them) a not necessarily benign force which will arise to replace the species and eliminate the human soul.

Luzi identifies Alan Turing as the true inventor of AI because, writes Luzi, Turing “pushed to its logical conclusion the calculability of thought” (p.34) With the “Turing machine” of 1936 Turing introduced the notion of the algorithm, a system which enables a  mechanical analysis and calculation to be carried out which is indistinguishable from the processing of data by the human brain. It was such algorithms which famously enabled Deep Blue to beat Garry Kasparov at chess and in recent years the term has entered common parlance as AI tools become more usual and widespread.

One of Elon Musk’s favorite authors, at least according to Luzi’s Martian, is Nick Bostrom, former director of the Future of Humanity Institute at Oxford University. Bostrom repeats the mantra according to Luzi’s Martian that

the sense and purpose of humanity is to accede to technological maturity…in other words to reach a level of  maturity corresponding to capacities which permit a level of economic productivity and control over nature  close to the highest level which could ever be reached….The ultimate end of industrialization is thus to become “master and owner of nature” (René Descartes)…In conformity with your cyborg religion and the systematic degradation of natural gifts (including human natural gifts) such progression to technological maturity is for you an ascendancy to divinity. To reach a point of such divinity means that existential risks are incurred which put humanity in jeopardy, they include the abusive utilization of nanotechnologies (malevolent nanobots), a nuclear holocaust, the apocalyptic diffusion of a biologically modified pathogen or the calamitously defective programming of a super-intelligence etc. The picture is more terrifying than that of Dante’s Inferno, but for Bostrom as for you, there is no choice: either humanity reaches a necessary level of technological maturity or humanity must needs perish. (pp 14-15)

Luzi argues that the impulse and purpose of AI is to increase control over human beings and to bring the war against nature to a successful conclusion by completely destroying all natural environments and replacing them with a virtual reality. He is very aware that AI is a product of evolution and reminds the reader that DARPA (Defense Advanced Research Projects Agency, a branch of the US Defense Department) defines its own aim as to “push the boundaries of scientific and technological innovation.”  The manner in which DARPA utilizes AI gives credence to Luzi’s argument that AI began as a military initiative and now plays a crucial role in warfare technology, aiming to reach an ever more efficient system of controlling, dominating and destroying. For Luzi, the beginnings of AI in warfare influences it in everything it does.  Turing himself is best know for an achievement in war: namely for breaking the German enigma code in World War II. The email, notes Luzi, began as a communication tool for the military. If most people are asked to think of AI, they might think of drones before they think of google.

Luzi writes:

The Second World War was a period of intense technological innovations, among them the computer. Cybernetics was born at the heart of the National Defense Research Committee as a result of the work of Norbert Wiener who perfected aerial defense, which led him to work on the optimization of creating a human-machine interface. (p.47)

Luzi quotes Paul Edwards from The Closed World: Computers and the Politics of Discourse to underline the close relationship of computer and then AI development to the military industrial complex and presents that complex in those very terms, an alliance of the armed forces and technology, especially computer technology.

The National Security Commission of Artificial Intelligence instituted by Congress in May 2018, includes among its standing members representatives of Amazon Web Services, Oracle, Microsoft Research Laboratories and Google Cloud…The commission has made recommendations aiming at furthering the progress of the development of artificial intelligence, automatic learning procedures and associated technologies with the purpose of responding globally to the security and defense needs of the United States.

Luzi believes that the evolutionary course which technology has taken and the link (or should one say interface?) between economic profit, artificial intelligence and the military means that states themselves have lost effective control of the dynamics of war and peace. “The global commitment to economic profit and military power makes it impossible for industrialized nations to end war and the technological preparations for it.” (p 56).

Of the Ukrainian war and the use of drones, Luzi writes: “the battalions of drones are dedicated to exceed the capacities of their human commanders, bringing us closer to the robot general.” (p.53)

Luzi believes that advocates of AI are advocates of an accelerated process of human evolution, writing,

[F]or the neo-Darwinian, Daniel Dennett, the evolution of living beings follows a blind and mechanical procedure and the human being is a robot composed of “millions of macro-molecular machines”. Similarly, NASA defines a living being as a self-maintaining system capable of Darwinian evolution. Maybe one day such a definition of a machine being will lead to the belief that a bibot, created from muscular cells controlled by electrical impulses, is a living being. (pp 33-34)

Maybe it will. Dennett famously insisted that competence precedes understanding and that the superiority or conscious cultural ability of Homo sapiens only lies in the hierarchical overview which people enjoy of what they have achieved (by contrast with the unawareness of termites, which individually, argues Dennett, work in way similar to human brain cells).  Luzi also writes about James Lovelock, an early “Green” activist and probably best known for his “Gaia” theory.  Luzi underlines what he sees as a contradiction in Lovelock’s understanding of technology:

Lovelock’s idealization of science alternates between asserting the necessity of dealing with the flux of precise information on all aspects of the world by means of a giant computer and the assertion that technology is neutral and that it is quite possible to only utilize those aspects of technology which are beneficial. What is not found in Lovecraft’s works is an acknowledgment of the political and ecological implications of a cybernetic complex, linking State, <use of large case S and no definite article is in the original>, capital and techno-science; the incompatibility of ethics and the obligation imposed upon citizens today to accept all technological “progress”; problems posed by replacement of human reflection by an automatic retroactive process. In short, an awareness of the deep antagonism between industrial development and the persistence of humanity. (pp 41-42)

These are strong words. They show two things: firstly, that for Luzi the fundamental conflict here is between the evolutionists and the anti-evolutionists. It is certainly striking that all keen advocates of AI laud evolution and profess unwavering atheism.  Secondly, Luzi considers a large part of Green politics to be a fraud, a Trojan horse. James Lovelock, who was passionate about “saving humankind” , argues in Gaia that different life forms interconnected with non-life forms on Earth constitute one whole, a living organism. His work appeared shortly after the publication of the Club of Rome’s Limits to Growth, ostensibly a manifesto against relentless progress, and can be seen to complement it. What is less well known, but Luzi reminds us, is that Lovelock worked as a consultant for NASA and for MI5, British Military Intelligence. He was also outspoken in his support for nuclear energy. Luzi is strongly implying that many ecologists are wolves in sheep’s clothing, and it is true that most Greens today appear remarkably nonchalant about Artificial Intelligence and advanced technical solutions to modern challenges: smart phones, solar panels and the like they regard and indeed promote as “environmentally friendly”.

The implication here and throughout the book seems to be that the real conflict of view is between those who regard nature as the cornerstone of value and those who look to science for their lodestar in life. Whatever else one may think of Luzi’s writing, the conflict of natural versus technical is an important paradigm to consider, one refreshingly different from the usual “liberal versus conservative”, or “globalist versus nationalist” or “elite versus populist” or “populist versus democratic” paradigm, which nearly all analysts fall back on when commenting on world events.

Like so much of his book, Luzi’s nature versus artificiality paradigm can give rise to a debate which would go far beyond the brief comments and observations which Luzi himself makes. His work can be said to practice “nudging”.  “Nudging” is a common feature of AI itself, namely in the field of advertising, whereby  purchasing habits online trigger algorithmic calculations to create suggestions and advertising banners which “nudge” the purchaser to buy or use again in the same reference area.  (“Youtube thinks you will like this”). Similarly,  Jacques Luzi’s book “nudges” the reader in the direction of many deeply interesting issues, without discussing them in any depth itself.

The “peaceful” uses of AI are not spared Luzi’s distrust and scorn.  He believes that the non-military application of AI has its own hidden agenda:

In the 1990’s, with a good dose of government aid, the widespread availability of the home computer led to the transition to commercial applications monopolized by Gafam, -viz. Google, Amazon, Facebook, Apple, Microsoft- in addition to the companies for whom the military is a habitual client: IBM, Intel, etc. The widespread availability of small individual computers opened the way to a neo-liberal reorganization of the financial markets. (pp 48-49).

Just as in military action, so in a neo-liberal world of buying and selling, notes Luzi, speed is decisive. The drive towards ever faster and more accurate AI in business, notably financial markets is obvious and is pushing AI to greater levels of achievement. The second level of AI, Artificial General Intelligence, AGI. is not very far away. Surprisingly, in his critique of what he superbly describes as a “simultaneously comfortable and brutal dehumanization” brought about by AI (p.106) Luzi does not examine the brutalizing effects of AI induced stress, especially on the old. The more that questions and problems and tasks are addressed by AI “solutions”, the less tolerant or patient an administration or authority is likely to be of  hesitant, slow human reactions, be they in the area of online buying and selling, banking, registration, communication, identity control, mobile Aps, QR codes and more. AI can be stressful for users, and because artificial, frequently counter-intuitive.  It is in the nature of AI that the slow moving or thinking humans will tend to be disadvantaged.

Luzi writes fervently on what he sees as the war of AI against nature. He quotes approvingly the French novelist Jean Giono from the novel Le Poids du ciel, published in 1938: “(Technical) civilization seeks to dominate nature, to restrict it, to compel it and ultimately to fight against it, to conquer it and in its secret heart, unbeknown even to itself, to destroy it.”  (p.61)

The lines are clearly drawn between those like Luzi, whose sensibility is religious, with a notion of the sacredness of nature and those on the other hand who belong to a tradition of dyspathy in the face of  natural order and who constantly want to improve or circumnavigate nature and “the natural course of events”. Thus Francis Bacon in Luzi’s emotive words “wished to effect the desacralization of nature by means of the sacralization of technics.” (p. 59). Human beings are now predisposed to accept AI solutions because most human beings are already living in an artificial environment. Since 2008, for the first time in human history, a majority of the world’s human population lives in cities, separated from a direct relationship with the natural environment and absorbing only at second hand much of its experience of what the natural world is like, not only at second hand, but increasingly with the assistance of Artificial Intelligence.

Luzi links Artificial Intelligence to industrial global capitalism, which he says, necessarily seeks the destruction of “inefficient”  autonomous subsistence economies. As AI increasingly takes over the tasks formerly carried out using natural intelligence, the human brain, Luzi believes, will atrophy and people be “reduced to what Zbigniew Brzezinski, Trilateral Commissioner  and advisor to President Carter, termed tittytainment, a Roman type mixture of bread, electronic games and video seduction, aimed at canalizing the boredom of a restless yet obsolete Homo industrialis.” (p. 83)

Luzi cites the ominous words of Lewis Mumford writing in The Myth of the Machine. If anyone has inspired Luzi, it is surely Mumford, and by the end of the book many readers will realize that What Artificial Intelligence Can’t Do is not a postscript to the book with the similar title by Drefus but to Mumford’s warnings about the mega-machine. Artificial Intelligence is the glue, so Luzi, being used to combine the expanding power of science with increasing human idiocy and passivity under the auspices of a ruthless ultra competent and efficient state, whose symbol is the all seeing eye, the Eye of Providence. Luzi writes ironically:

Technology, Prosperity and Power should be engraved into all the public and private monuments of the proto-states later to become fully industrialized so that their true Gods might be celebrated with dignity. Technology: increase the power to transform by great leaps forward natural material to hand (including human material)  transformed into a calculable machine, removable and reconfigurable. Prosperity: organize the abundance of artifacts created by the quest for power.. Power: employ power and status and extend them to ensure technological and therefore military superiority. (pp 91-92)

Luzi quotes James Scott to the effect that we can now see the advent of an artificial society not created by the force of custom or the accidents of history but one following strictly scientific and rational criteria. It comes as no surprise to the reader that Luzi mentions Samuel Bentham’s “panoptica” the eye that places prison inmates under 24 hour surveillance.

Luzi’s book concludes with a dialogue between the writer and “a creator of Artificial Intelligence” presumably fictional, although this is not made clear. The creator greets the writer with the words “hello scribbler” and that sets the tone for their “discussion” which is an exchange of insults rather than a meaningful dialogue and leaves Luzi’s readers where they started: our species is confronted with a dark future, one dominated by a totalitarian machine dedicated to war, the destruction of the natural world and the crushing of human freedom with the long term end of replacing human beings with machines. By taking the uncompromisingly anti-scientific stand which it does, this work offers no “solution” to what it sees as a remorseless evolution towards the extinction of Homo sapiens.

Can those who wish to halt the progression of AI do so? It seems unlikely that Artificial Intelligence is something which could be halted in its tracks, for two reasons: one is that what is known to man is difficult to banish, especially in a world of global communication. A scenario in which everyone abandons Artificial Intelligence out of fear of its potential is barely conceivable, because if only a small minority can still use it, it will still evolve and grow in power, with the rest of the population defenseless against the minority. Secondly, AI has its own evolutionary momentum and an evolutionary momentum is naturally, ironical as that word sounds is in this context, unstoppable until a greater more efficient or more devastating force or more successful competitor opposes it. An end to Artificial Intelligence would necessitate a political and economic reaction amounting to a global cataclysm. Perhaps Jacques Luzi will write a book explaining how he envisages such a cataclysm. That said, his critique in What Artificial Intelligence Can’t Do is far from foolish. The enthusiasm, one can fairly say blind enthusiasm, that many have for AI and disregard of its obvious potential abuse by the power hungry and authoritarians of every hue, AI’s evident potential to dehumanize, is sufficient reason  to hope that AI can be restrained or channeled. Clearly, combating the perils of AI is the subject for another book, but one obvious point can be briefly made here. Luzi does not make it however, either because it is just too obvious, or because he is too immersed in his uncompromising invective even to consider it.

The prerequisite for individual freedom in the face of fast developing AI is surely this: a cast iron guarantee that the use of AI solutions do not become mandatory or exclusive solutions. AI as much as possible should be prevented from becoming the sole option in a procedure. That means for example, that electronic payment systems should never be accepted as the only accepted means of payment.  It means that electronic voting is never adopted as the only possible way to vote, it means that smart cities, smart homes and the rest do not leave citizens with no choice but to accept them. From that point of view, the debate will not be about what AI can or can’t do, but rather about a struggle to ensure that AI will not be permitted to replace the old ways of doing things entirely, to ensure that the old ways will always matter and never be entirely replaced.