Non-Robots and Their World

Another book on my list for this quarter was The Mechanical Mind in History, edited by Philip Husbands, Owen Holland, and Michael Wheeler. The book is a collection of articles that are mostly about early thinkers in cognitive science and artificial intelligence. Most of the articles seem to be written by people in the field for others in the field, and a lot of the assumptions of the field are shared (e.g. that it is unproblematic and good to think of the mind as a machine, part of what “we” are about). So, for example, there is an article about Descartes that shows how surprisingly his ideas are in concert with ideas in artificial intelligence, so we should like Descartes. So most of these articles are of limited use to me right now.

There are some exceptions. The article on Charles Babbage’s difference engine and the ideas about intelligence that he formed along with it was extremely interesting. It was fascinating to see how much of the artificial intelligence project was already being pursued by Babbage during that time. It is also useful to see how the idea of mechanizing the mind, and viewing the mind as a mechanism, goes back to the time when mechanization was such a powerful economic imperative. The historical detail in this article will be useful to return to, as will its bibliographical references.

The other article I found useful in the book is one I had encountered before but hadn’t fully understood, which is Hubert Dreyfus’s piece titled, “Why Heideggerian AI Failed and How Fixing It Would Require Making It More Heideggerian.” In this article Dreyfus summarizes the problems he has with “good old fashioned AI” (GOFAI), which is based on the processing of represented knowledge, addresses three responses to GOFAI along Heideggerian lines that he thinks are inadequte (including Phil Agre’s), and talks about what he thinks is a potentially successful direction for AI based on a neurodynamic model of how the brain thinks. I’m getting quite a bit of mileage out of this piece in terms of my own thinking about my interests. What Dreyfus is talking about is attempts to develop a strong AI program that creates real minds like ours, rather than the weak AI sort of program that is now where most of the research money is going, since weak AI programs develop useful applications. But I think that phenomenological objections to strong AI programs are actually relevant to the technologies developed by weak AI programs, for the reason that these technologies interact with real people and do so using models of how people’s minds work, in order to be helpfully predictive (that is, helpfully for whoever deploys the technology). Whether these models are of minds that function on the basis of reasoning through representations or by interacting with their worlds dynamically is relevant to the ultimate design and functioning of the technologies. I can see my dissertation possibly being concerned with that question in particular, with respect to some specific technology, like maybe an expert system of some kind, or some kind of an interface.

There is another thing that Dreyfus’s article got me thinking about that may help me sort out some issues for my work. It’s something I’ve thought about before but need to clarify further, and it has to do with an ambiguity in much of the work in artificial intelligence and cognitive science, which is that they’re often unclear on whether they are talking about how minds work in the broadest sense, covering a broad range of mentation and consciousness, or whether they are talking about logical reasoning, the emergence of which from a broader dynamic system is I think a very interesting thing. The broader sense of mind is about how we live, or as Dreyfus might say, how we cope, but I would also add, how we enjoy. The more specific case of reasoning or logic or intelligence does seem to involve representations (and not unproblematically as Wittgenstein and others have shown). But also of interest about it is that reasoning is concerned not so much with living or coping as with knowing truth. Nietzscheans would say that amounts to nothing, but I wouldn’t dismiss it so easily. I think knowledge of truth is something the meaning of which can only be understood with reference to metaphysics, and on that account presents a further ontological problem for researchers in AI, though it’s one that tends to lead to a stalemate. I think this is the case even if we accept that judgments we make about what is true are at the level of representational language processing that is emergent from dynamic life-processes, because what is emergent isn’t necessarily reduceable to the phenomena at the level from which it emerges. Most importantly, the formation of subjectivity, and of the sense of meaning that guides our lives and makes life worth living, run counter to the condition that people sometimes experience as mechanistic modules in a mechanistic social process, a condition that is engendered by belief in mechanical minds and interaction with systems that treat us mechanically.

I re-read the section of Ron Day’s book where he discusses Heidegger and a Heideggerian corrective to a reified concept of information, and I found it a good expression of my core standpoint, and found it a lot clearer in its expression than I did when I first read it. That’s all I’ll say about it at present, except to comment that it puzzles me that Heidegger was a Nazi and that this is supposed to taint his philosophy so badly, when at the same time it seems that a major implication of what he wrote would be to deplore the way the Nazi state turned society into a machine in which people were simply resources.

I read Philosophers and Machines, a collection of articles on this history of technology taken from a journal called Isis and edited by Otto Mayr. The book was published in 1976, before the concept of “information” became so central. There wasn’t much in the book about information technology or much that read “information” into technological processes historically. In fact it almost seemed odd the way information was not mentioned. Even in the article on the influence of Maxwell’s work on governors on the development of cybernetics did not talk about it in terms of information, but rather in terms of dynamics. The articles were not really very philosophical, really more like straight histories of specific technologies, sometimes a little speculative. The most interesting article was not actually about a technology but about an Italian poet who wrote about the technologies of the industrial age, Leopardi. He was attacking what he evidently called “romantic” poetry, but what he meant was poetry that found something inspiring and romantic in these new technologies. The interesting thing in the article, and the thing that could be valuable as a bibliographic resource, is a section on automata going back to the Greeks. The author speculates that a device Leopardi writes about was a clever automaton created for entertainment purposes. No real commentary on what the existence of those things meant, or what Leopardi’s reaction to them meant. So while interesting, the book was not really helpful for my purposes this quarter.

I also read Elting Morison’s little book, Men, Machines and Modern Times, which is the transcript of a series of lectures that definitely have the feeling of speech. This book was published in 1966, and although a couple of the chapters have to do with computers, the concept of information is not present in this book either. The book was ahead of its time in certain ways though. Morison gave a lot of attention to things that sociologists of technology beginning in the 80s would be concerned with – the nature of social dynamics around the introduction of new technologies – how they are developed, adopted, spread, and the effects that they have. His essays were pretty interesting, but again, not especially helpful for my purposes this quarter for the most part. There were some passages that did give me some good insights. His discussion of bureaucracy in one chapter made me see that a lot of my concerns about information technology actually have to do with the bureaucratic ordering of life that it perpetuates. Morison’s commentary on bureaucracy was interesting and useful, and I will probably come back to it. Unfortunately, the book doesn’t have bibliographic references.

I put the Beniger aside without finishing it, because I felt that I had gotten his point and got what I needed to out of the book. But it has stayed on my mind, and I think I need to continue reading it, albeit from a different angle. Beniger reifies information from the start, and his account of the history of information leading up to the information age does not historicize this reification (the way Ron Day’s book does). I put the book down partly out of annoyance about that. But now I am thinking that reading it from a different angle could help show how his reification of information entered into culture as a result of the changes he describes. In order to help me get a better sense of what information meant prior to this reification, I am re-reading Ron Day’s section on Heideggerian critique of the metaphysics of information. I also can’t help thinking of Walter Ong in this connection, and I’m wondering if perhaps I will end up working with his ideas.

James R. Beniger. The Control Revolution: Technological and Economic Origins of the Information Society.

This book helped me clarify my thinking about some important issues, in forcing me to confront what I didn’t like about it.

To summarize the book, Beniger understands information along information-theory lines, as a reduction of entropy. Central to his view of the history of information society is the idea that all organisms are organisms because they are organized for the purpose of countering entropy, so that they can maintain the energy to continue to exist. All activity is explained in terms of countering entropy and maintaining the organism’s existence. This is the case for individual people and for social groups. Everything is working in concert within an organism in order to take energy from outside of itself so that it can continue living. Furthermore, also central to his view is that organisms do this through programs that determine their actions. He is explicitly deterministic and anti-metaphysical. Beniger uses these ideas – avoiding heat death and behaving according to programs in order to do so, to explain all phenomena stemming from organic life going back to the first unicellular organisms and continuing up through microprocessors and genetic engineering. Information, for Beniger, is the ordering of matter and energy for the purpose of controlling a system’s behavior in an organized effort to prevent heat death. I realize I am making it sound funny, but it doesn’t take any effort. On top of this wild reach, Beniger uses somewhat insulting language when he discusses opposing views.

The book does have some interesting analytical categories and concepts. There are three main aspects of organisms that he wants to look at in this light: their Being (maintaining existence counter to entropy), their Behavior (using programs to deal with events), and their Becoming (self-reprogramming in order to deal with longer-term changing conditions). Being, Behavior, and Becoming seem like grand, all-encompassing categories, but given the meaninglessness of life in Beniger’s account they hardly seem adequate to cover the meaning that life actually does have for us. These three aspects have static and dynamic sides to them. The static and dynamic sides of Being are organization and metabolism; the static and dynamic sides of Becoming are responsiveness and adaptability; and the static and dynamic sides of Becoming are reproduction and selection. These biological concepts are fundamental for him because he believes that every emergent phenomena can be fully explained in reductive terms, and as he sees it all technology is an extension of our physiological capacities.

I can acknowledge that synthesizing all of this is pretty clever and even pretty grand, but there is so much that it sadly does not account for, in a phenomenological sense especially.

That is why it is a useful book for clarifying the importance of certain things for me philosophically. His account of life is indeed what you get without metaphysics, and shows, for me, the need for metaphysics. It helps me to see that my own existential concerns relating to people and machines have a connection to metaphysics, in that there is something about being a person (or perhaps just being a being) that has a metaphysical aspect to it. Existentially, our subjectivity is all important (the Kierkegaardian understanding of subjectivity). In terms of philosophy outside existentialism, this has the implication that metaphysics is important, for explaining subjectivity in objective terms that are not reductive.

Beniger’s book is still somewhat useful as an economic history if you reject his ideas. The path from the industrial revolution to the computer revolution is important. Beniger’s concept of control is not without application there, either, so far as it goes. But reading him is a pain given his extremely bleak view of things. I’m glad to have made use of his book for clarifying purposes.

I read Lewis Mumford’s small 1952 book Art and Technics, find it a good starting point for thinking and rich in ideas, but lacking in useful citations. He turned out to be just a little less romantic about the gap between our subjective, art-needing nature and our technologies than I thought at first. He gave quite a bit of attention to the value of what he called “machine arts”: printing and photography. Essential to both of these for him is that their ability to reproduce their materials democratizes art. He also talked about the way their mechanical basis contributes its own aesthetic based on function, and advocates simplicity over ornamentation in these forms. But he is also very worried about the effects of the ease of mechanical reproduction, in that it creates a situation where artworks can become neutralized by overexposure, and we lose the ability to appreciate the qualitative as we are pulled along by the mechanical capacity to create quantity out of standardized forms and processes. He introduces some concepts that can serve as ideas regarding essential aspects of technology, which is something that strongly went out of fashion in the 80s with the emergence of the sociology of technology. But it seems to me that reproduceability based on standardization is an essential and problem-laden aspect of much technology and something that we shouldn’t overlook. It does seem to be the source of the concerns for subjectivity, soul, nature, etc. in the romantic era responses to industrialization. It relates to information studies in the way that these technological trends led to formal representations of ideas suitable for manipulation by computing machines, and that this rendered the world of thought into a technological order. This happened after the time that Mumford was writing, for the most part, yet his concerns apply to it. His book gives me some expressions of concerns that I already have had, but not any real theoretical anchors.

I am doing another 596 (directed study) this quarter. It is directed at the same concerns as last quarter’s, but from a different angle. This quarter I will be reading works on the history of technology and information technology, with the aim of gaining a more historical understanding of the concept of information as it has been technologized.

I am starting with Lewis Mumford’s Art and Technics. It’s a polemical little book that is less about history and more about the modern transition to a technological mode of life. It is a bit romantic, but it does articulate some real concerns about the role of technology in our culture from a humanist perspective. It is important to state exactly what those concerns are so that what one says is more than a lyrical expression of romanticism that could be suspected as nostalgia for the womb. I am thinking about that problem as I read his book.

The readings I am planning beyond this are:

James Beniger, The Control Revolution

Elting Morison, Men, Machines and Modern Times

James Gleick, The Information

Noah Kennedy, The Industrialization of Intelligence

Otto Mayr, editor, Philosophers and Machines

Otto Mayr, Authority, Liberty & Automatic Machinery in Early Modern Europe

Husbands, Holland, Wheeler, eds., The Mechanical Mind in History

Paul N. Edwards, The Closed World: Computers and the Politics of Discourse in Cold War America

I will comment on these books as I read them….

It’s now the 2nd week of spring quarter. I’m taking Dr. Blanchette’s class in academic work, which should be very helpful, and Chris Kelty’s class in the political economy of information, which is not quite what I would like from a class on the topic, but should be fairly interesting. For a third class, I planned on taking a class on Marx, Weber and Durkheim taught in the sociology department, but it turned out the focus of the class was different from what I wanted. It is about constructing theories in sociology from data. So I unenrolled. Now I am planning to do a 596 on the history of information technology, because it is something I want to learn about and won’t have an opportunity to do so otherwise. I haven’t talked to Leah about the idea yet, but I am sure it will be fine.

In Kelty’s class, the first week was theory, and that meant learning his epistemological standpoint for research. We read the preface to Nietzsche’s Genealogy of Morals, a piece by Foucault reflecting on genealogy as a method, and a chapter from a book by John Dewey on logic, which expressed his situated pragmatic perspective. Class discussion yesterday was at a highly theoretical level that left a lot of the class behind, but I have a few classmates who really rock when it comes to philosophy. It was pretty enjoyable. But the emphasis on the class is on “participation,” particularly instances of participation in various kinds of online communities for doing different things. This is the subject of Kelty’s NSF-funded research. I am not entirely sure what I think about the project. I raised a couple of critical questions about it in class, and I hope I was not to contrary. I think it was fine, because I don’t personally feel that I am against it, so I think my questions probably seemed to be what they were, i.e. questions. At any rate, I am enjoying the class so far and appreciate Kelty’s erudition and knowledge of philosophy. It should be interesting.

And something else. The course and Kelty’s research may be important to me for a dissertation project that I have thought of and am now a little obsessed with. It is something different from what I have been concerned about intellectually. It is not about the effect of formal systems on our interaction with culture. It is kind of a relief to have something to think about that is more concrete. The idea is about a way of letting ordinary people get their messages into physical public space through electronic signs. These would be, at the big scale, giant electronic billboards, and at the smallest scale, three-foot wide LED signs in one color that would be placed inside a public/private space or in the window of a business. In one model, people access time on the sign by bidding for one hour or 24 hour slots, possibly sharing the time with someone else. If the sign is in a business, the message rotates with messages provided by the business, in exchange for the use of their space. Also rotating into the sign would be something like, “Public voices in public space: www.this-sign.net”, or just the URL, which is fairly self-explanatory. (I have already registered the domain.) It occurred to me to put a big sign up on my building in Duluth, which is in a highly-trafficked, visible location, but there are probably prohibitive zoning issues. I will probably do the first pilot study with a small sign in my brother’s bar. My friend Jill has a good location in Milwaukee where we could build a free-standing digital sign at a later stage, and she is interested. Ok, so I have mentioned the commercial, bidding model, but for the purposes of the study (and civic goodness), there would be another model operating, which would be to let an online community determine what messages to put on the sign, either as a Sunday project for signs that have commercial messages, or as the focus of a particular project, like for instance an outdoor electronic message board somewhere along Bruin Walk on the UCLA campus. The study could be modeled after Kelty’s and feed into it. His research is partly ethnographic but also uses variables to facilitate comparison. It’s quite nice. This project could use data going a little deeper, because sources of data would be not only a description of the project within Kelty’s framework of description, but also the messages that go on the sign, website statistics, the discussion in the online community itself, comments on the website, and local press coverage. I spent the weekend talking to people about the idea and got a lot of useful feeback. The only faculty member I have had a chance to talk to about it yet is Greg Leazer. He found it to be a little bit of a crazy idea, because of the uncertainty of actually getting people to participate given how people tend to tune out advertising, and he has a point. However, a sign along the Bruin Walk was his suggestion, and it is a good one. Today I will talk to Johanna about it, and Leah on Thursday. I think a logical committee would consist of Leah, Chris Kelty, Doug Kellner, and possibly someone else. I really hope this works, because the concreteness is feeling like a relief after I’ve been struggling to figure out what to do with a more fundamentally philosophical concern.

Part of my coursework this quarter was an independent study in AI, as I mentioned earlier. This post is a wrap-up where I reflect on how my thinking has developed through the readings I did for this independent study.

I started the readings having already had some exposure to work on philosophical issues around AI, especially Hubert Dreyfus’s work. I had some issues in mind, which are the issues I have thought about throughout the quarter.

I read Mortimer Taube’s Computers and Common Sense, Nils J. Nilsson’s The Quest for Artificial Intelligence: A History of Ideas and Achievements, Ernest Nagel and James. R. Newman’s Gödel’s Proof, John Haugeland, ed., Mind Design II: Philosophy, Psychology, Artificial Intelligence, J.C.R. Licklider’s paper titled, “Man-Computer Symbiosis,” and H. M. Collins’ Collins, Artificial Experts: Social Knowledge and Machines. I also did some reading for Dr. Lievrouw’s class on the public sphere in working on a literature review on the debate around Habermas’s views on technology that apply to my issues. These included, among other things, articles by Marcuse, Habermas, Andrew Feenberg, Tyler Veak, and Robert Pippin about the question of whether technology itself carries its own ideology with respect to our relationship to reason and nature.

Taube’s book, which I started with, focused my attention on the question of formalization and its implications. I explored a particular logical problem in formal systems in the book on Gödel’s proof, with consideration of its implications for AI. Formalization of knowledge is at issue in modernity, technology, research methods in the social sciences, epistemology generally, and, clearly, information studies. It was good to begin articulating this focus and to work on it throughout the quarter.

Nilsson’s The Quest for Artificial Intelligence opened my mind a bit to approaches to artificial intelligence that differ from the ideas that I have always had about it, particularly connectionist, situated, and embodied AIs. These approaches partially deal with two major types of philosophical objection: objections having to do with the limitations of logical knowledge representation (logical incompleteness and non-exhaustivity of rules) and phenomenological objections having to do with situatedness and embodiment as preconditions for real thinking. The main result of this in my own thinking was to realize that my concerns about AI have to do with a number of issues that are actually distinct, though related. I began to break down my interests in AI into a number of issues, intellectual frameworks for approaching them, and areas of application that might work as topics of research. I will continue to build on and refine these lists as I clarify the problems for myself.

Reading Mind Design II, which is a collection of classic articles, was a breakthrough for me, because it gave me a much more thorough exposure to the philosophical debate over AI and its problems than I had had before. I was exposed to a number of perspectives and to work in cognitive science where AI theorists have been working on problems posed by philosophers. This book helped me gain the background that I hoped to get out of the independent study. I developed my own responses to a number of the articles there, including Alan Turing’s seminal piece where he introduced what came to be known as the “Turing Test.” I was very glad to be reacquainted with John Searle’s “Chinese Room” argument here. The debate over connectionist approaches and their implications was interesting. Connectionist theory is essentially a theory of minds as networks, which dovetails into Cassells’ work. Much of the work in here helped me think further about questions around formalization.

Reading Licklider’s paper got me thinking about the connection, or lack thereof, between these discussions about AI and work in HCI. I will need to become familiar with the HCI field, but I am pretty sure the issues I want to deal with have not been addressed there. That is not to say that they couldn’t be, which may mean that HCI is an possible arena for my work. I should look at the HCI literature with an eye for research that could use a response in terms of the issues that I am concerned with.

Finally, H.M. Collins’ Artificial Experts helped me think further about the question of formalization, through its investigation of what exactly formalization is, in social terms. Collins is not critical in the way I want to be, but his work will be very useful to me. More than anything else I have read, this book deals directly with my concerns about formalization. (Ron Day’s book also relates very closely.)

Coming out of this I have the start to a new reading list. I want to read, really read this time, Phil Agre’s book. I want to read Winograd and Flores’ Computers and Cognition, because it is important, although I don’t feel a great need for more of the philosophical work from the 80s. I want to read Jerry Fodor’s more recent book, The Mind Doesn’t Work That Way. I want to read some Lakoff. I seriously need to read Sherry Turkle. I want to read Pamela McCorduck’s Machines Who Think. And I want to read Lucille Suchman’s Plans and Situated Actions (and I will look into the possibility of taking a class from her at Irvine). I also want to find some work that addresses issues of modernity and technology, but particularly w/respect to the process of formalization, from an existentialist perspective, which I have not quite found. Most of what I have found that criticizes technology from a critical theory or phenomenological perspective has mostly to do with instrumental reason and the turning of nature into a resource, rather than with formalization, digitalization, etc. I should talk to Greg about this. I may use this list for another 596, considering advice. Maybe I should do the 596 with Greg, if he has the time.

I have done a little exploring of the website Chatbots.org, which is a pretty big resource for people who are interested in chatbots. A lot of the work on AI that I have been reading has been about early chatbots as demonstrations of ideas in AI. now chatbots are a pretty big industry, mostly used by companies who want to provide an interactive type of help without paying for actual customer service employees. Basically knowledge base interfaces in place of extensive (or not so extensive) FAQ’s. These kind of chat bots don’t need to be able to talk about anything the user wants to talk about, because they can say they are there for a certain purpose if they don’t understand something. But they are still interesting as applications of AI.

On the site I discovered that there are a lot of services that allow you to create your own chatbots, often for free. I think I will explore some of these. Examples are Pandorabot, IMified, and AIPioneer Chatbot System. It will be interesting to see how flexible these are. I could see possibly using one of these to do an experiment, comparing people’s experience with their chat partner along some parameter.

H.M. Collins, Artificial Experts: Social Knowledge and Machines. MIT Press, 1990.

Harry Collins wrote this book about AI issues from the position of the sociology of science and technology. He was responding to the discussion from the 80s that has been my primary interest in this independent study: Dreyfus, Haugeland, Searle, etc.

The book will be extremely useful to me going forward, because it provides insight into my basic motivating problems from a perspective that is outside that of the philosophers that I have been reading. While Collins is critical of the basic claims made by people in expert systems and AI, he doesn’t accept the philosophers’ à priori style of argument regarding them, seeing those arguments as beside the point in terms of how computers can be made to interact with us socially and function in our projects with us. Although his main interest is the nature of machine intelligence as such and how it differs from what we usually think of as “human intelligence,” that interest turns out to be an interest in a particular way of thinking that people, particularly in the West, have invented for ourselves. As he formulates it, the dichotomy at issue is between behavior and action, and then concerted behavior and concerted action. Behavior is understandable purely from the outside, through objective measurement and description, though Collins, characteristically for a sociologist of science, emphasizes (and this turns out to be really important) that the measurements we make are all dependent on cultural agreements about how things can be measured and classified. Critical to his argument is a concept of digitization, whereby we take, for example, a myriad of variations in shape and agree that they are all variations on the letter “s,” which is a digital object, discrete and meaningful as a sign or token rather than materially. This digitization is an essential part of how we create a kind of action he calls behavior-specific actions, where certain behaviors (physical expressions) can have only one meaning in the context. This is machine-like behavior according to formalized rules. These rules enable this kind of action to be transfered to machines. Most of the book is concerned with the limits of machine-like behavior and the behavior of thinking machines in a social context.

There are a lot of tools and concepts in the book that I will find useful, as well as useful reponses to some of the philosophical statements that I have been feeling at home with, but in the end I find the book oddly uncritical. I think this is because he gravitates toward a network theory of society that lacks the kind of normative conceptions that I expect. He is not particularly worried about what our culture’s use of machine-like action may mean, even as he looks ahead to its increasingly sophisticated application. I find that a little odd.

Although I took extensive notes on this book, I am not going to be able to comment on it in depth at this point in the academic quarter. I think this is sufficient for communicating this gist of this book and its meaning for me right now. I can add that some new thoughts it is giving me in a practical sense come from the empirical side of the book, which is made up of a number of demonstrations he created using relatively easy to use software for programming expert systems and playing with AI interfaces. Thinking about things I have read about, I think perhaps some practical experimentation could be a part of my dissertation.