I’m reading a book about superintelligence, and naturally the thing that has stayed with me so far is the owl.
Nick Bostrom opens Superintelligence with a fable about sparrows considering bringing an owl into their lives. An owl could do things they cannot. Most of the sparrows are enthusiastic. A smaller number wonder whether it might be sensible to work out how to control an owl before inviting one in.
It is a neat way into a much bigger problem, although it also helps that I have a weakness for owls. I’m a Wednesdayite, so they come with the territory, and some of my friends have occasionally described me as “Owly”, whatever that means. Put an owl at the beginning of a book about artificial intelligence and apparently you have my attention.
That is probably as good a place as any to start some reading notes.
I’ve been getting back into reading properly. Not reading because I ought to, or because a particular book is new, but following books that connect with questions I already seem to be carrying around.
Except, looking at this particular group, “what I’m reading” is not quite accurate. Some are books I am reading now. Others I first encountered years ago and have found myself returning to. What I hadn’t really noticed until I started writing this is how connected that journey has been.
These are not reviews. They are notes about what stuck, what has come back and why some old ideas suddenly seem worth looking at again.
Superintelligence is not a new book, which is partly what makes reading it now interesting.
The owl story lands differently when artificial intelligence is no longer something I encounter mainly through speculation about the future. I use it. I build things with it, ask it questions and use it to interrogate ideas. Increasingly, I see it appearing inside ordinary work.
That does not mean we have arrived at Bostrom’s superintelligence. We plainly haven’t. But it changes the emotional distance from the argument. The sparrows are no longer debating whether owls might theoretically be useful. There are feathers around the place.
What I like about the metaphor is that it keeps the question simple. The interesting part is not just what a more capable intelligence can do, but what happens to the relationship once something much more capable is inside the system.
I suspect the owl will stay with me longer than many of the technical details.
Critical Path is a reread, although enough time has passed that I suspect quite a lot of it will feel new again.
I first read it a long time ago and remember being struck by Buckminster Fuller himself as much as by any particular argument. His work and designs seemed genuinely extraordinary to me: somebody moving between architecture, engineering, resources, economics and society without much respect for the boundaries between them.
There are fragments that have somehow survived in my memory. His ideas about value and resources are in there somewhere, along with a stubbornly vague recollection involving cows. I could try to reconstruct what that was supposed to mean, but rereading the book seems a better option.
What certainly survived was the scale of the thinking. Fuller was willing to look at the systems humanity had built and ask whether their apparent permanence was really justified. That made an impression on me then and probably explains some of my continuing attraction to systems thinking now.
It is also why I reached for the book again.
I spend much more of my working life thinking explicitly about systems these days: organisations, institutions, technology, incentives and the constraints people inherit without necessarily choosing them. AI has now added something that was not part of my thinking when I first read Fuller. I’m curious whether advances in artificial intelligence change the practicality of any of the systems he imagined.
Some things that were previously too complicated, information-heavy or difficult to coordinate may no longer be so. That does not make Fuller right, and technology does not make the human parts of a system disappear. If anything, experience has made me more suspicious of solutions that work beautifully until people arrive.
But it does make the book worth another look.
I read The Sovereign Individual much more recently and found it fascinating. It is one of those books that leaves you with a few ideas that keep resurfacing afterwards.
The one I remember most clearly is what the authors describe as the returns to violence. What I took from it was the idea that the structure of society is influenced by who can exercise coercive power effectively and what they gain from being able to do so. If technology changes that calculation — particularly by allowing individuals to protect themselves, move resources or exercise forms of power that once belonged overwhelmingly to states — then it potentially changes the value of the institutions built around that power too.
It made me think about Elon Musk. Not because the book somehow predicts Musk, but because he is such a visible example of how much capability can now accumulate around one individual. Companies, enormous wealth, communications infrastructure, spacecraft, artificial intelligence and political influence make for a combination of power that would once have been much more closely associated with states.
The other idea that made an impression was the suggestion that politics is reaching the later stages of a roughly five-hundred-year cycle, with the information age beginning to undermine institutions developed for a previous technological era.
It is a big hypothesis, but I thought it was a decent one.
There is certainly something recognisable in the idea that institutions can remain in place long after the conditions that made them effective have begun to change. Politics continues, elections happen and governments govern, but there also seems to be a widespread sense that parts of the system no longer work particularly well and perhaps are not especially well suited to the world developing around them.
Whether that really amounts to the end of a five-hundred-year political cycle is a much bigger claim. I don’t need to be certain of that for the idea to be useful. What stayed with me was the possibility that some of what looks like political dysfunction may actually be the strain of institutions designed for one technological age trying to operate in another.
That was probably the strongest impression the book left on me: technology does not just give individuals and institutions new tools. At some point it can alter the balance of power that made those institutions make sense in the first place.
This one takes me much further back.
I read The Zeitgeist Movement Defined when the movement was relatively fresh and, at the time, I also had an interest in the Green Party.
How times change.
What interested me then was not so much a particular political programme as the idea of looking at society itself as a system. I was interested in whether some of the problems we repeatedly attempt to solve individually might actually be consequences of how the wider system is designed.
Environmental pressures, inequality, resources, incentives, technology and human behaviour look rather different when you stop treating them as completely separate subjects.
I found the possibility of thinking outside the usual boundaries exciting. I still do, although I am probably less susceptible now to believing that identifying the system automatically means you have found the solution.
People rather complicate things.
They carry history, loyalties, ambitions, families, fears, beliefs, status and very different ideas of what a good society should actually produce. A model can make perfect sense and still fail because the people inside it do not behave in the way the model requires.
What I had forgotten until recently was how important the book was in sending me elsewhere. My interest in those larger societal systems led me towards Buckminster Fuller and eventually Critical Path. From there came other books, other ideas and other questions.
Looking back, the reading journey feels considerably less random than it did while I was taking it.
I don’t think I have ever consciously designed a reading programme around systems, power, institutions and technology. One book has simply tended to leave something behind that makes another one interesting. Sometimes the connection happens immediately; sometimes an idea sits there for years until something changes and an old book suddenly seems worth taking off the shelf again.
AI seems to be doing that to quite a few of them.
Ideas about how we allocate resources, organise society, distribute power or coordinate complicated systems feel different when intelligence itself is becoming cheaper and more widely available. Predictions made twenty or thirty years ago can look remarkably prescient in one paragraph and wildly optimistic in the next. Books I read much earlier now encounter a version of me with different experience and different questions.
That is probably what I am enjoying most about reading again. I am not particularly looking for authors who will confirm what I already think, and I’m increasingly comfortable finishing a book without deciding whether I agree with it.
Some books are useful because they change your mind. Others are useful because they give you something to argue with for the next twenty years.
And every now and again, apparently, one has an owl in it.