On the plane to Ethiopia, and while sweating in my tent at night from the relentless heat at Filoha, I read Turing’s Cathedral, a book my Dad told me I should read. He was right, it’s well worth reading. It’s also timely, as my wife had us watch the movie Deus_ex_Machina for a recent family movie night.
In Deus_ex_Machina, a lone genius constructs an artificial intelligence (AI), which he then subjects to a Turing test to see if she can pass as a conscious being. The movie is thought provoking and disturbing. You can have a brief conversation with the AI, who is named Ava, here. She can even draw a picture of you.
The book Turing’s Cathedral describes the building of one of the first artificial intelligences: the MANIAC computer at Princeton’s Institute for Advanced Studies in the 1940s and 50s. The building of the MANIAC was an intensely collaborative team project. While Hollywood loves isolated mad scientists working in remote lairs, nowadays real scientific advances are usually the result of team effort. No one person can have enough expertise to do everything that needs to be done. AIs that can pass the Turing test will be the result of teamwork on a massive scale.
The book is titled Turing’s Cathedral, but the main character is not Alan Turing (who provided the theoretical foundations for computers), but Johnny von Neumann, who among other things, invented game theory and cellular automata, and played key roles in developing the first atomic weapons and digital computers. Von Neumann was an intensely social genius who built and led the team of people who made the MANIAC, and who traveled extensively, providing intellectual connectivity between Princeton, Los Alamos, Cambridge and other key hives of activity.
The author of Turing’s Cathedral is George Dyson, who grew up at the Institute for Advanced Studies, where his father Freeman Dyson was a fellow. Dyson shows how the development of computers and atomic weapons were intimately linked, and also connected to evolution and genetics. MANIAC’s first job was to run simulations of thermonuclear explosions. Soon after, Nils Aall Barricelli programmed MANIAC to run simulations of evolution.
Evolution, it turns out, is key to thinking about logic, mathematics, and the stuff of thought. One of the central puzzles of mathematics in the early 20th Century was whether all true mathematical statements could be derived from a simple set of axioms. Mathematician David Hilbert argued that “from a strictly limited set of axioms, all mathematical truths could be reached by a sequence of well-defined logical steps.” (Dyson 2012: 49) Bertrand Russell and Alfred North Whitehead tried to do this in Principia Mathematica, which despite covering nearly 2,000 pages “still left fundamental questions unresolved” (Dyson 2012: 49). Von Neumann took on the challenge in 1925 in a paper, “An axiomatization of set theory,” which provided a more concise and more nearly complete answer to the problem. But in 1936, Kurt Gödel published a paper that argued that the project could never be completed, because mathematics was fundamentally incomplete: “within any formal system sufficiently powerful to include ordinary arithmetic, there will always be undecidable statements that cannot be proved true, yet cannot be proved false” (Dyson 2012:50).
This incompleteness of mathematics, and logical systems generally, relates to insight and intuition. Systematic and logical thinking alone can only get you so far. Insights involve leaps, making unexpected connections. Insight is evolutionarily equivalent to mutation: random changes that sometimes result in improvements.
When Turing worked on the Manchester Mark I computer in 1949, he made sure to install a random number generator. This enabled the computer to take advantage of mutations, or “guesses,” and learn from its mistakes.
Intellectual thought, and biological evolution, both depend on two key things: accumulated change, and mutation. Accumulated change ensures that past advances are preserved. Mutation provides the material for new solutions to problems. Most mutations don’t work, but some do. Adding the good mutations to the accumulated wisdom of previous generations permits advances to be made much more rapidly than if everything had to be attempted at once.
During World War II, Turing developed a machine to break the code used in German war communications. The code was too complicated to be solved all at once. But by using a mutating, evolving process, Turing managed to evolve solutions to the problems.
Computers evolve, and DNA is a computer. When Dad told me I needed to read Turing’s Cathedral, he told me, “DNA is a Turing machine!” And he’s right. Dyson doesn’t really develop this point, but he hints at it.
A Turing machine is a universal computer. When Turing proposed the machine in 1936, many people believed that thought was somehow distinct from machines: thoughts are the work of souls, which machines can never have. But Turing proposed that a machine could produce all computable statements. Turing described the machine as having a tape fed into it. The machine can read symbols on the tape, write symbols on the tape, and move the tape left or right. The machine is mindless, but it produces intelligent behavior. Our world today is densely populated with Turing machines running on our computers, phones, and the cloud.
DNA is essentially the tape in a cellular Turing machine. It stores digital information to run the programs that the cell carries out. The cellular mechanisms read the DNA tape, copying portions of it into RNA, which may either be translated into proteins, or used to regulate particular portions of the DNA, or other cellular processes. DNA is mindless, but it acts intelligently. DNA encodes complicated subroutines that involve numerous genes that regulate the production of proteins. But because DNA is digital and universal, in principle it could be used to store any kind of information. Perhaps someday artificial computers will use DNA memories.
For the first few hundred pages, I thought that Turing’s cathedral was the MANIAC computer. But no, as it turns out, the cathedral is Google: that vast conglomeration of Turing machines that knows so much about everyone who has ever searched the Internet. Ava, the AI in Deus_ex_Machina, is created by a man who runs a company, Blue Book, that is a thinly veiled version of Google. Though Ava is created by a lone genius, she learns to understand humans by being granted access to Blue Book searches and cell phone conversations. In this way the AI is created by all of us. And Dyson argues that Google and other online giants, such as Facebook and Amazon, are evolving towards artificial intelligence as they learn from us.
And as computers become more like us, we are increasingly becoming like computers. Reasonably well off people in rich countries have all become cyborgs, and this trend is expanding worldwide. I started becoming a cyborg when Dad gave me an IBM clone computer. This was not my first computer, but it was the first one with a floppy drive (instead of a cassette tape) and enough memory to store what I wrote: papers for classes, college application essays, stories, notes for science fiction worlds, and other personal writings. Soon a substantial portion of what I think of as me was stored on this machine. It didn’t have as much memory as my brain, but it remembered more accurately.
My current laptop retains most of these old memories (though some have been lost through decay, copying errors, failures to store things properly, and loss of compatibility between old files and new programs). But my cyborg existence doesn’t depend on a single machine anymore. Instead it is smeared out across the Cloud: Gmail, Facebook, Dropbox, Snapfish, Google’s memory of my search history, Amazon’s memory of my purchases. The Web may know more about me than I do. With every bit we post, with every search we enter, with every product we buy, we feeding the Web information about human behavior and preferences. We are all building the next AI. And as the Web (or Google or whatever) evolves awareness – well, I hope it is benign.
Dyson, G. (2012). Turing’s Cathedral: The Origins of the Digital Universe, Pantheon.