A Box of a Trillion Souls

pandora's box

“The cybernetic structure of a person has been refined by a very large, very long, and very deep encounter with physical reality.”                                                                          

Jaron Lanier

 

Stephen Wolfram may, or may not, have a justifiable reputation for intellectual egotism, but I like him anyway. I am pretty sure this is because, whenever I listen to the man speak I most often  walk away no so much with answers as a whole new way to frame questions I had never seen before, but sometimes I’m just left mesmerized, or perhaps bewildered, by an image he’s managed to draw.

A while back during a talk/demo of at the SXSW festival he managed to do this when he brought up the idea of “a box of a trillion souls”. He didn’t elaborate much, but left it there, after which I chewed on the metaphor for a few days and then returned to real life, which can be mesmerizing and bewildering enough.

A couple days ago I finally came across an explanation of the idea in a speech by Wolfram over at John Brockman’s Edge.org  There, Wolfram also opined on the near future of computation and the place of  humanity in the universe. I’ll cover those thoughts first before I get to his box full of souls.

One of the things I like about Wolfram is that, uncommonly for a technologist, he tends to approach explanations historically. In his speech he lays out a sort of history of information that begins with information being conveyed genetically with the emergence of life, moves to the interplay between individual and environment with the development of more complex life, and flowers in spoken language with the appearance of humans.

Spoken language eventually gave rise to the written word, though it took almost all of human history for writing to become nearly as common as speaking. For most of that time reading and writing were monopolized by elites. A good deal of mathematics, as well has moved from being utilized by an intellectual minority to being part of the furniture of the everyday world, though more advanced maths continues to be understandable by specialists alone.

The next stage in Wolfram’s history of information, the one we are living in, is the age of code. What distinguishes code from language is that it is “immediately executable” by which I understand him to mean that code is not just some set of instructions but, when run, the thing those instruction describe itself.

Much like reading, writing and basic mathematics before the invention of printing and universal education, code is today largely understood by specialists only. Yet rather than endure for millennia, as was the case with the monopoly of writing by the clerisy, Wolfram sees the age of non-universal code to be ending almost as soon as it began.

Wolfram believes that specialized computer languages will soon give way to “natural language programming”.  A fully developed form of natural language programming would be readable by both computers and human beings- numbers of people far beyond those who know how to code, so that code would be written in typical human languages like English or Chinese. He is not just making idle predictions, but has created a free program that allows you to play around with his own version of a NLP.

Wolfram makes some predictions as to what a world where natural language programming became ubiquitous- where just as many people could code as could now write- might look like. The gap between law and code would largely disappear. The vast majority of people, including school children, would have at the ability to program computers to do interesting things, including perform original research. As computers become embedded in objects the environment itself will be open to the programming of everyone.

All this would seem very good for us humans and would be even better given that Wolfram sees it as the prelude to the end of scarcity, including the scarcity of time that we now call death. But then comes the AI. Artificial intelligence will be both the necessary tool to explore the possibility space of the computational universe and the primary intelligence via which we interact with the entirety of the realm of human thought.  Yet at some threshold AI might leave us with nothing to do as it will have become the best and most efficient way to meet our goals.

What makes Wolfram nervous isn’t human extinction at the hands of super-intelligence so much as what becomes of us after scarcity and death have been eliminated and AI can achieve any goal- artistic ones included- better than us. This is Wolfram’s  vision of the not too far off future, which given the competition with even current reality, isn’t near sufficiently weird enough. It’s only when he starts speculating on where this whole thing is ultimately headed that anything so strange as Boltzmann brains make their appearance, yet something like them does and no one should be surprised given his ideas about the nature of computation.

One of Wolfram’s most intriguing, and controversial, ideas is something he calls computational equivalence. With this idea he claims not only that computation is ubiquitous across nature, but that the line between intelligence and merely complicated behavior that grows out of ubiquitous natural computation is exceedingly difficult to draw.

For Wolfram the colloquialism that “the weather has a mind of its own” isn’t just a way of complaining that the rain has ruined your picnic, but, in an almost panpsychic or pantheistic way, captures a deeper truth that natural phenomenon are the enactment of a sort of algorithm, which, he would claim, is why we can successfully model their behavior with other algorithms we call computer “simulations.” The word simulations needs quotes because, if I understand him, Wolfram is claiming that there would be no difference between a computer simulation of something at a certain level of description and the real thing.

It’s this view of computation that leads Wolfram to his far future and his box of a trillion souls. For if there is no difference between a perfect simulation and reality, if there is nothing that will prevent us from creating perfect simulations, at some point in the future however far off, then it makes perfect sense to think that some digitized version of you, which as far as you are concerned will be you, could end up in a “box”, along with billions or trillions of similar digitized persons, with perhaps millions or more copies of  you.   

I’ve tried to figure out where exactly this conclusion for an idea I otherwise find attractive, that is computational equivalence, goes wrong other just in terms of my intuition or common sense. I think the problem might come down to the fact that while many complex phenomenon in nature may have computer like features, they are not universal Turing machines i.e. general purpose computers, but machines whose information processing is very limited and specific to that established by its makeup.

Natural systems, including animals like ourselves, are more like the Tic-Tac-Toe machine built by the young Danny Hillis and described in his excellent primer on computers, that is still insightful decades after its publication- The Pattern on the Stone. Of course, animals such as ourselves can show vastly more types of behavior and exhibit a form of freedom of a totally different order than a game tree built out of circuit boards and lightbulbs, but, much like such a specialized machine, the way in which we think isn’t a form of generalized computation, but shows a definitive shape based on our evolutionary, cultural and personal history. In a way, Wolfram’s overgeneralization of computational equivalence negates what I find to be his as or more important idea of the central importance of particular pasts in defining who we are as a species, people and individuals.

Oddly enough, Wolfram falls into the exact same trap that the science-fiction writer Stanislaw Lem fell into after he had hit upon an equally intriguing, though in some ways quite opposite understanding of computation and information.

Lem believed that the whole system of computation and mathematics human beings use to describe the world was a kind of historical artifact for which there much be much better alternatives buried in the way systems that had evolved over time processed information. A key scientific task he thought would be to uncover this natural computation and find ways to use it in the way we now use math and computation.

Where this leads him is to precisely the same conclusion as Wolfram, the possibility of building a actual world in the form of simulation. He imagines the future designers of just such simulated worlds:

“Imagine that our Designer now wants to turn his world into a habitat for intelligent beings. What would present the greatest difficulty here? Preventing them from dying right away? No, this condition is taken for granted. His main difficulty lies in ensuring that the creatures for whom the Universe will serve as a habitat do not find out about its “artificiality”. One is right to be concerned that the very suspicion that there may be something else beyond “everything” would immediately encourage them to seek exit from this “everything” considering themselves prisoners of the latter, they would storm their surroundings, looking for a way out- out of pure curiosity- if nothing else.

…We must not therefore cover up or barricade the exit. We must make its existence impossible to guess.” ( 291 -292)

Yet it seems to me that moving from the idea that things in the world: a storm, the structure of a sea-shell, the way particular types of problems are solved are algorithmic to the conclusion that the entirety of the world could be hung together in one universal  algorithm is a massive overgeneralization. Perhaps there is some sense that the universe might be said to be weakly analogous, not to one program, but to a computer language (the laws of physics) upon which an infinite ensemble of other programs can be instantiated, but which is structured so as to make some programs more likely to be run while deeming others impossible. Nevertheless, which programs actually get executed is subject to some degree of contingency- all that happens in the universe is not determined from initial conditions. Our choices actually count.

Still, such a view continues to treat the question of corporal structure as irrelevant, whereas structure itself may be primary.

The idea of the world as code, or DNA as a sort of code is incredibly attractive because it implies a kind of plasticity which equals power. What gets lost however, is something of the artifact like nature of everything that is, the physical stuff that surrounds us, life, our cultural environment. All that is exists as the product of a unique history where every moment counts, and this history, as it were, is the anchor that determines what is real. Asserting the world is or could be fully represented as a simulation either implies that such a simulation possesses the kinds of compression and abstraction, along with the ahistorical plasticity that comes with mathematics and code or it doesn’t, and if it doesn’t, it’s difficult to say how anything like a person, let alone, trillions of persons, or a universe could actually, rather than merely symbolically, be contained in a box even a beautiful one.

For the truly real can perhaps most often be identified by its refusal to be abstracted away or compressed and by its stubborn resistance to our desire to give it whatever shape we please.

 

Advertisements

Summa Technologiae, or why the trouble with science is religion

Soviet Space Art 2

Before I read Lee Billings’ piece in the fall issue of Nautilus, I had no idea that in addition to being one of the world’s greatest science-fiction writers, Stanislaw Lem had written what became a forgotten book, a tome that was intended to be the overarching text of the technological age his 1966 Summa Technologiae.

I won’t go into detail on Billings’ thought provoking piece, suffice it to say that he leads us to question whether we have lost something of Lem’s depth with our current batch of Silicon Valley singularitarians who have largely repackaged ideas first fleshed out by the Polish novelist. Billings also leads us to wonder whether our focus on the either fantastic or terrifying aspects of the future are causing us to forget the human suffering that is here, right now, at our feet. I encourage you to check the piece out for yourself. In addition to Billings there’s also an excellent review of the Summa Technologiae by Giulio Prisco, here.

Rather than look at either Billings’ or Prisco’s piece , I will try to lay out some of the ideas found in Lem’s 1966 Summa Technologiae a book at once dense almost to the point of incomprehensibility, yet full of insights we should pay attention to as the world Lem imagines unfolds before our eyes, or at least seems to be doing so for some of us.

The first thing that stuck me when reading the Summa Technologiae was that it wasn’t our version of Aquinas’ Summa Theologica from which Lem got his tract’s name. In the 13th century Summa Theologica you find the voice of a speaker supremely confident in both the rationality of the world and the confidence that he understands it. Aquinas, of course, didn’t really possess such a comprehensive understanding, but it is perhaps odd that the more we have learned the more confused we have become, and Lem’s Summa Technologiae reflects some of this modern confusion.

Unlike Aquinas, Lem is in a sense blind to our destination, and what he is trying to do is to probe into the blackness of the future to sense the contours of the ultimate fate of our scientific and our technological civilization. Lem seeks to identify the roadblocks we likely will encounter if we are to continue our technological advancement- roadblocks that are important to identify because we have yet to find any evidence in the form of extraterrestrial civilizations that they can be actually be overcome.

The fundamental aspect of technological advancement is that it has become both its own reward and a trap. We have become absolutely dependent on scientific and technological progress as long as population growth continues- for if technological advancement stumbles and population continues to increase living standards would precipitously fall.

The problem Lem sees is that science is growing faster than the population, and in order to keep up with it we would eventually have to turn all human beings into scientists, and then some. Science advances by exploring the whole of the possibility space – we can’t predict which of its explorations will produce something useful in advance, or which avenues will prove fruitful in terms of our understanding.  It’s as if the territory has become so large we at some point will no longer have enough people to explore all of it, and thus will have to narrow the number of regions we look at. This narrowing puts us at risk of not finding the keys to El Dorado, so to speak, because we will not have asked and answered the right questions. We are approaching what Lem calls “the information peak.”

The absolutist nature of the scientific endeavor itself, our need to explore all avenues or risk losing something essential, for Lem, will inevitably lead to our attempt to create artificial intelligence. We will pursue AI to act as what he calls an “intelligence amplifier” though Lem is thinking of AI in a whole new way where computational processes mimic those done in nature, like the physics “calculations” of a tennis genius like Roger Federer, or my 4 year old learning how to throw a football.

Lem through the power of his imagination alone seemed to anticipate both some of the problems we would encounter when trying to build AI, and the ways we would likely try to escape them. For all their seeming intelligence our machines lack the behavioral complexity of even lower animals, let alone human intelligence, and one of the main roads away from these limitations is getting silicon intelligence to be more like that of carbon based creatures – not even so much as “brain like” as “biological like”.

Way back in the 1960’s, Lem thought we would need to learn from biological systems if we wanted to really get to something like artificial intelligence- think, for example, of how much more bang you get for your buck when you contrast DNA and a computer program. A computer program get you some interesting or useful behavior or process done by machine, DNA, well… it get you programmers.

The somewhat uncomfortable fact about designing machine intelligence around biological like processes is that they might end up a lot like how the human brain works- a process largely invisible to its possessor. How did I catch that ball? Damned if I know, or damned if I know if one is asking what was the internal process that led me to catch the ball.

Just going about our way in the world we make “calculations” that would make the world’s fastest supercomputers green with envy, were they actually sophisticated enough to experience envy. We do all the incredible things we do without having any solid idea, either scientific or internal, about how it is we are doing them. Lem thinks “real” AI will be like that. It will be able to out think us because it will be a species of natural intelligence like our own, and just like our own thinking, we will soon become hard pressed to explain how exactly it arrived at some conclusion or decision. Truly intelligent AI will end up being a “black box”.

Our increasingly complex societies might need such AI’s to serve the role of what Lem calls “Homostats”- machines that run the complex interactions of society. The dilemma appears the minute we surrender the responsibility to make our decisions to a homostat. For then the possibility opens that we will not be able to know how a homostat arrived at its decision, or what a homostat is actually trying to accomplish when it informs us that we should do something, or even, what goal lies behind its actions.

It’s quite a fascinating view, that science might be epistemologically insatiable in this way, and that, at some point it will grow beyond the limits of human intelligence, either our sheer numbers, or our mental capacity, and that the only way out of this which still includes technological progress will be to develop “naturalistic” AI: that very soon our societies will be so complicated that they will require the use of such AIs to manage them.

I am not sure if the view is right, but to my eyes at least it’s got much more meat on its bones than current singularitarian arguments about “exponential trends” that take little account of the fact, as Lem does, that at least one outcome is that the scientific wave we’ve been riding for five or so centuries will run into a wall we will find impossible to crest.

Yet perhaps the most intriguing ideas in Lem’s Summa Technologiae are those imaginative leaps that he throws at the reader almost as an aside, with little reference to his overall theory of technological development. Take his metaphor of the mathematician as a sort of crazy  of “tailor”.

He makes clothes but does not know for whom. He does not think about it. Some of his clothes are spherical without any opening for legs or feet…

The tailor is only concerned with one thing: he wants them to be consistent.

He takes his clothes to a massive warehouse. If we could enter it, we would discover clothes that could fit an octopus, others fit trees, butterflies, or people.

The great majority of his clothes would not find any application. (171-172)

This is Lem’s clever way of explaining the so-called “unreasonable effectiveness of mathematics” a view that is the opposite of current day platonists such as Max Tegmark who holds all mathematical structures to be real even if we are unable to find actual examples of them in our universe.

Lem thinks math is more like a ladder. It allows you to climb high enough to see a house, or even a mountain, but shouldn’t be confused with the house or the mountain itself. Indeed, most of the time, as his tailor example is meant to show, the ladder mathematics builds isn’t good for climbing at all. This is why Lem thinks we will need to learn “nature’s language” rather than go on using our invented language of mathematics if we want to continue to progress.

For all its originality and freshness, the Summa Technologiae is not without its problems. Once we start imagining that we can play the role of creator it seems we are unable to escape the same moral failings the religious would have once held against God. Here is Lem imagining a far future when we could create a simulated universe inhabited by virtual people who think they are real.

Imagine that our Designer now wants to turn his world into a habitat for intelligent beings. What would present the greatest difficulty here? Preventing them from dying right away? No, this condition is taken for granted. His main difficulty lies in ensuring that the creatures for whom the Universe will serve as a habitat do not find out about its “artificiality”. One is right to be concerned that the very suspicion that there may be something else beyond “everything” would immediately encourage them to seek exit from this “everything” considering themselves prisoners of the latter, they would storm their surroundings, looking for a way out- out of pure curiosity- if nothing else.

…We must not therefore cover up or barricade the exit. We must make its existence impossible to guess. ( 291 -292)

If Lem is ultimately proven correct, and we arrive at this destination where we create virtual universes with sentient inhabitants whom we keep blind to their true nature, then science will have ended where it began- with the demon imagined by Descartes.

The scientific revolution commenced when it was realized that we could neither trust our own sense nor our traditions to tell us the truth about the world – the most famous example of which was the discovery that the earth, contrary to all perception and history, traveled around the sun and not the other way round. The first generation of scientists who emerged in a world in which God had “hidden his face” couldn’t help but understand this new view of nature as the creator’s elaborate puzzle that we would have to painfully reconstruct, piece by piece, hidden as it was beneath the illusion of our own “fallen” senses and the false post-edenic world we had built around them.

Yet a curious new fear arises with this: What if the creator had designed the world so that it could never be understood? Descartes, at the very beginning of science, reconceptualized the creator as an omnipotent demon.

I will suppose then not that Deity who is sovereignly good and the fountain of truth but that some malignant demon who is at once exceedingly potent and deceitful has employed all his artifice to deceive me I will suppose that the sky the air the earth colours figures sounds and all external things are nothing better than the illusions of dreams by means of which this being has laid snares for my credulity.

Descartes’ escape from this dreaded absence of intelligibility was his famous “cogito ergo sum”, the certainty a reasoning being has in its own existence. The entire world could be an illusion, but the fact of one’s own consciousness was nothing that not even an all powerful demon would be able to take away.

What Lem’s resurrection of the demon imagined by Descartes tells us is just how deeply religious thinking still lies at the heart of science. The idea has become secularized, and part of our mythology of science-fiction, but its still there, indeed, its the only scientifically fashionable form of creationism around. As proof, not even the most secular among us unlikely bat an eye at experiments to test whether the universe is an “infinite hologram”. And if such experiments show fruit they will either point to a designer that allowed us to know our reality or didn’t care to “bar the exits”, but the crazy thing, if one takes Lem and Descartes seriously, is that their creator/demon is ultimately as ineffable and unrouteable as the old ideas of God from which it descended. For any failure to prove the hypothesis that we are living in a “simulation” can be brushed aside on the basis that whatever has brought about this simulation doesn’t really want us to know. It’s only a short step from there to unraveling the whole truth concept at the heart of science. Like any garden variety creationists we end up seeing the proof’s of science as part of God’s (or whatever we’re now calling God) infinitely clever ruse.

The idea that there might be an unseeable creator behind it all is just one of the religious myths buried deeply in science, a myth that traces its origins less from the day-to-day mundane experiments and theory building of actual scientists than from a certain type of scientific philosophy or science-fiction that has constructed a cosmology around what science is for and what science means. It is the mythology the singularitarians and others who followed Lem remain trapped in often to the detriment of both technology and science. What is a shame is that these are myths that Lem, even with his expansive powers of imagination, did not dream widely enough to see beyond.