What I've Been Reading

17 October 2018

  • Noam Chomsky: Syntactic Structures

Since computational linguistics will become one of my main research topics over the next few months I wanted to look into this locus classicus. Chomsky’s great mind shows itself in the programmatic nature of this text on transformational grammar. One could see someone else defend many of the conclusions, for example arguing that focussing on phrase structure alone falls short of providing a sufficiently strong grammar for natural languages such as English, while lacking Chomsky’s grasp on how this fits in a more general approach to the study of language and even human cognition in general. Of course, Chomsky has moved on multiple times since the publication of this book, but his programmatic approach, his capacity to address specific issues with a whole theory of his field in mind, has remained the same. Chomsky’s theories offer plenty to disagree with, but to match his overall program remains a challenge.

  • Nils Nilsson: The Quest for Artificial Intelligence: A History of Ideas and Achievements

To be honest, I had expected a different text. Although Nilsson has been part of AI research for many decades, his book on the history of AI does not indulge in personal anecdotes. No juicy stories about infighting and rivalries. Apparently always the proper researcher, Nilsson’s book reads at time like an extended survey article. The picture he paints of AI research is one of continuous progress. Certain approaches fail, or at least stop to receive as much attention as they used to, and there are breakthroughs, but all of these remain adjustements and part of an overall trajectory. At this point, I wouldn’t dare to criticise Nilsson’s portrayal of the history of AI. It also supports my scepticism about the assumption that general, (super-)human-level AI is now finally waiting just around the corner. Neural networks have been there from the beginning and while we have certainly made progress, from the perspective of this book, they have been only a few steps forward towards the ultimate goal of AI research.

  • Steven Pinker: The Language Instinct

One of my courses at Cambridge suggested reading Pinker’s book in preparation for computational linguistics. It does contain a few very important technical details anyone in this field should be familiar with, but I assume that the ideas was more to convey a general mindset about how to approach language. It succeeded in my case and the book deserves to be named amongst other outstanding texts for the educated public. I just wish Pinker wouldn’t be so damn overconfident when he leaves his main field of research. Last time I picked up one of his books, I put it away after thirty pages or so. His political digressions annoyed me immensely, not so much the content but the unjustifiably self-assured tone with which it was presented. And Pinker lacks the great style and self-irony of the late Jerry Fodor, who mixed his diatribe against relativism with one against fiberglass powerboats (in a passage which Pinker quotes). That being said, in this earlier book Pinker managed to contain himself and his attack on the marvens of school grammar hits the target much better.

  • Daniel C. Dennett: Consciousness Explained

Consciousness Explained is Dennett at his best. It’s his magnum opus. Nowhere else has Dennett come closer to convincing me of his illusinionist approach to consciousness - and during my research visits at Tufts, I sat in on one of his courses. Still, I find myself unable to buy into it, instead I end up in limbo: neither accepting the mysterious approaches to consciousness (no epiphenomenalism, no panpsychism, or any odd dualism), nor Dennett’s solution of explaining consciousness away (that’s basically what he is doing, as I heard him grant in person). There are philosophical concerns I have and then there are these pesky intuitions. Interestingly, both conflict zones tend to circle around Dennett’s not-quite-metaphysics, in particular the role of fiction. Dennett doesn’t really want to get into the metaphysics of fictions and for a good reason. It is messy, difficult, and might appear like a major distraction from the main topic: consciousness. But then again, if consciousness becomes a fiction of sorts, then you have to come clean what fictions of sorts are. Dennett’s paper “Real Patterns” takes a step or two towards answering this question, but it falls far short both of filling the philosophical gaps and of making me comfortable.

  • Daisy Johnson: Everything Under

A book that calls for being spoiled. Read the end and wonder whether it is intellectually self-indulgent, a brilliant way to pull the threads together, or what the reader (should have) expected. At least for two reasons authors are constantly tempted to start writing about language: First, it allows them to pull all kinds of neat tricks, mixing medium and message. Second, language is something the authors feel comfortable with. Johnson makes language a theme, but she does so subtly enough for it not to become obnoxious. I’m for philosophical reasons annoyed by the thought that our language limits our understanding of the world. Of course it does, to some extent in some sense, but not in the way and to the extent people stumbling upon the thought for the first time tend to believe. For people with another background,