Michael Anissimov/Gotterdammerung Review
And here we have a disorganized review/response to the "Gotterdammerung: Dawn of the Human-Made Gods" book by Bill Hibbard. The book is online at http://www.ssec.wisc.edu/~billh/g/ -- Michael Anissimov
Gotterdammerung Review, by Michael Anissimov
“Gotterdammerung:Dawn of the Human-made Gods”, renamed “Super-Intelligent Machines” for publication, is a lengthy introduction to speculation regarding the future of artificial intelligence, plus much more. This summary and review is a fairly in-depth examination and critique of Hibbard’s introductory method and conclusions, so a bit of background knowledge on this subject may be required to appreciate the importance of the argued points and their sovereign impact on humanity’s near-term future. The tricky thing about speculation in these areas is that writers commonly tend to refer to the broad outlines of their predictions as truths, contradicting other futurist writers and creating an overall imbalanced intellectual environment among the pool of laypeople that speculate about the future of our planet and society. Part of the purpose of this summary and review is to contrast Hibbard’s ideas with another perspective, hopefully encouraging the reader to form a better mental pictures of the respective possibilities and their likelihoods, and gain a better perspective of the important subtopics in this fascinating and important quadrant of speculation.
Chapter 1 immediately invokes the metaphor of mythological deities, and comments on how science is slowly starting to answer all of our profound questions, putting religious folks in an odd position. The construction of machines with consciousness will turn typical human philosophies on their head most of all, because what we originally thought was so special and unique about humans alone will find another instantiation in complex arrangements of inanimate matter. But this is what we were all along, Hibbard says, and the benefits of these machines will be so great that we will rejoyce upon their arrival. Hibbard postulates that intelligent machines willl become so much smarter than human beings so quickly that humans will invoke religious feelings when trying to the bridge the gap of understanding. Human intelligence enhancement is not mentioned. An analogy is drawn between the future creation of superintelligent machines and historic theological thought. I believe this encourages the the reading audience to anthropomorphize - or theomorphize - the projected arrival of such entities.
Chapter 2 introduces arguments for the feasibility of machine intelligence, advantages machines might have over humans, a guess for the arrival of superintelligent machines, and the commonly considered issues of machine consciousness and emotion. I believe the prediction of 2100 for the arrival of superintelligent machines is radically conservative, based on the “intuitive linear” view of progress rather than the “historical exponential” trend which is actually taking place. (See Kurzweil’s Singularity is Near precis.) While the massive intelligence potential of these machines is addressed, the possibility of revising or completely rebuilding the default emotional structures that humans have isn’t really addressed. Hibbard mentions the critical difference that machines will have on our lives when they are able to understand our emotions - an understanding which would imply General Intelligence. While Hibbard would highlight emotional understanding, I would emphasize the increase of intelligence in general and label emotional understanding and communication as one of many benefits of this foundational intelligence.
Chapter 3 examines computers as tools. Hibbard points out that science and technology are simply shorthand for knowledge and tools, and that despite their unique connotations, people reap the benefits of advanced knowledge and tools on a daily basis. He comments on the robustness of the computing industry due to its widely useful nature and the Empirical Regularity of Moore’s Law. Being a computer programmer himself, Hibbard notes that his productivity in writing new programs has increased by an order of magnitude over the past 30 years. Computers are making people more efficient and offering new sorts of services that would have been difficult to imagine a few years ago. Hibbard imagines a time when computers are ubiquitous in society and this has major social, legal, economic, philosophical and personal implications, but this is where Hibbard’s projections of the future diverge from mine. I don’t believe that a society with ubiquitous computing could last very long without creating a greater-than-human intelligence, at which point most (or all) predictions will go up in smoke. Human legal and social systems tend to lag slightly behind technological advances, so I don’t think that these extra-technological changes will truly sink in before humanity breaks through the critical threshold. But envisioning a national or global community possessing ubuiquitous computing is an intellectually worthy thought experiment - and has valuable lessons to teach as far as technology-society interaction is concerned.
In a society where computing devices are semi-intelligent and embedded in many of the objects we use on a daily basis, technology would be more personal than most people ever would have expected. A weak sort of telepathy would develop in technologically advanced geographical regions, thanks to the overlapping benefits of easy, free long-distance communication, productivity-facilitating virtual environments and networks, instant high-quality information, personal assistant software, a variety of available interfaces, and many other mutually augmentive features. The critical point of interest here, however, is the transformation of computers from phone-TV-typewriter all-in-one devices to true extensions of human minds and personalities; more powerfully augmenting a greater and greater variety of specific human needs and concerns. When it begins to look genuinely silly to chide or abstain from using computer technology in larger and larger social and demographic sectors, public opinion will naturally become much more open to funding more computer research and accepting once-radical philosophical positions based on the recognition of technology’s reliable power. Computers will be less annoying, more socially accepted, easier to use, profoundly ubiquitous, and more impressive and magical as time goes on. “Economic crossover” will automate more and more jobs while the GDP and stock markets soar more and more reliably.
Hibbard, incorrectly in my opinion, claims that the increasing intelligence of computers will result in humans moving from traditional work lifestyles into hobby-centric modes of living. While better computing technology would eliminate much of the grunt work in society, I don’t think that this pushes us down a one-way track to academic overspecialization, overblown artistic fascination, or endless cocktail parties. While large groups could choose to forsake work for hobbies, (bearing in mind that I don’t forsee such a society could come into existence in the first place due to other technological inevitabilities which would eclipse all of this) economic impetuses and competitive trends will create novel, difficult-to-predict occupations which will be equally demanding, time and intelligence-wise, as many modern occupations. While being an insurance agent or banker is less physically demanding than working on a farm, they might require comparable time investments and a specific frame of mind. Instead of humans delegating more and more information-processing to autonomous computers and going off to play cyber-croquet, ambituous individuals will simply move themselves higher and higher up the hierarchial ladder of information manipulation, creation, and processing, performing so many useful tasks so proficiently in such a short period of time that the greatest entrepreneurs and researchers of today would be put to shame. If advanced ubiquitous computing emerged practically instantly, so that the First World’s memetic fabric had practically zero time to adjust, then I concede that large portions of society might initially work much less, but as new subcultures and communities spring up around never-before-possible productivity schemes, many individuals might actually work more and rest less, because not only will work bring in greater financial returns, but it will be easier and more enjoyable!
Since computers are currently programmed to primarily conduct cut-and-dry rational and procedural tasks, we anticipate that our emotions towards them in the indefinite future will inevitably follow suit. Hibbard insightfully recognizes that computers will soon shatter this stereotypical straightjacket and form emotional bonds with human beings, bonds potentially stronger than those between humans themselves. While he realistically conveys some of the impact that emotional bonds with machines will bring, he ignores the cognitive fact that machines this sophisticated will surely be able to think millions or billions of times faster than human beings, leaving open an opportunity for Recursive Self-Enhancement on timescales completely orthogonal to traditional human assumptions about minds, and an arrival so earth-shaking that it completely invalidates all but the least anthropocentric of premises. Notwithstanding the omission of this critical point, Hibbard points out the inevitable moral double standard and technology shock that many would experience as a result of human-equivalent machines, and hypothesizes that machine motivations will constitute pleasing human beings.
Chapter 4 presents the arguments of three prominent scientists and philosophers that machines cannot be intelligent. It discusses rebuttals to these arguments, as well as what they can teach us about the difficulties of actually building intelligent machines.
Chapter 5 discusses an overview of biological and medical research on human brains. Human brains provide our only example of a physical implementation of intelligence and consciousness, and hence are a valuable guide to understanding the nature of intelligent machines.
Chapter 6 describes a vision of intelligent machines and their role in human society, offering the author’s vision of how increasing machine intelligence could progress and impact a technologically sophisticated human civilization over the course of many decades.
Chapter 7 explores the morality of future machines and underscores the need that these machines be built as benevolent from the start. The Asimov Laws are introduced, and their inherent fallacies are uncovered, a step that many technologists and science fiction authors tend to ignore or brush aside. These laws are anthropomorphic, skipping over interactions between machines, and succeptible to the natural ambiguity of language, among many other things. Asimov Laws are appealing because they are simple and apparently concrete, although their simplicity and concreteness does more justice to a computer program than to superintelligent artificial intelligences. The only way that the integrity of machine entities can be ensured is by instilling them with a capacity for moral judgement and learning skills.
Chapter 8 explores the memetic issues of intelligent machines; namely, how will the masses handle it? How should they be educated regarding the emergence of machine intelligence? My first qualm for this chapter is that it assumes that the whole of society will play a role in educating and guiding these machines to a state of perfect unconditional love for all humans, but for better or for worse, I think that this is unlikely. The first machine mind created will likely be running at millions of times human thinking speed, and among other advantages, will give this entity the mental resources to quickly bootstrap into superintelligence, fundamentally changing the world before many humans have a chance to influence its’ developmental course. This cognitive fact about machine minds has deep philosophical and moral issues of its own, but I won’t go into them in detail here. Hibbard writes; “super-intelligent machines will be regulated by governments”, a sentence that comes across as odd to me as if I unearthed a Pleistocene relic with the words “homo sapiens sapiens will be regulated by administrative coalitions of homo sapiens habilises, when they evolve”. Machines of subhuman intelligence, yes, but once you reach true superhuman intelligence, the dynamic of the ballgame we’re all in will totally change.
Hibbard draws an analogy between the 21st century’s expanded definitions of “work” and the type of jobs that humans might take up after the arrival of super-intelligent machines, but I think that this analogy, while creative, may be a bit misleading. If superintelligent machines are capable of literally doing all the work, physical and intellectual, then that means that humans genuinely aren’t at the helm of the world’s progress anymore, and the time would have come to accept it. Any indignation we would feel at this fact, however powerful now, will quickly be calmed when these superintelligences assist us in boosting our own intelligence and rearranging our emotions so that our egocentric indignation dissappears. This will be part of the growing-up process for the entire human race.
