Daniel Dennett thinks a lot about things thinking about themselves thinking
Daniel Dennett, the Austin B. Fletcher Professor of Philosophy at Tufts University, is one of the most prominent modern philosophers and thinkers in the areas of cognition, the evolution of consciousness, and the development of the human mind, whatever that may be. A prolific author, he has published books discussing evolution (Darwin's Dangerous Idea: Evolution and the Meanings of Life), the roots of religious belief (Breaking the Spell: Religion as a Natural Phenomenon), cognitive strategies (Intuition Pumps and Other Tools for Thinking), and many other, related areas. Over his nearly 50 years of activity in this area he has returned time and again to fundamental questions such as "what is a mind" and "how did humans develop consciousness". In Bacteria to Bach and Back: The Evolution of Minds he presents a broad ranging, carefully considered summary of his argument that minds, consciousness and all the paraphernalia of our internal lives has developed in a logical, step-wise fashion depending only on natural selection, Darwinian evolution, and the eventual development of a persistent symbolic culture. In this thick, sweeping book Dennett covers many, many subjects, each in their own way relevant to the development of his central thesis, yet quite astonishing in their breadth and scope. A quick perusal of the topics listed in the index is enough to make your head spin: Beatrix Potter syndrome, Bayesian Hierarchical Coding, George Carlin jokes, the Beatles, Johann Sebastian Bach, William Shakespeare, the app Shazam, Shannon's information theory, Elon Musk, deep learning and GOFAI (Good Old-Fashioned Artificial Intelligence research), all these and more have a role to play at some point in the book.
Describing the pieces, then the board
The first third of the book is devoted to definitions, explanations, characterizations and general setting up of the playing field on which his argument for the development of modern minds is played out. Dennett revisits Darwinian evolutionary theory and the concept of natural selection, and a quick overview of the neo-Darwinian Modern Synthesis that combines natural selection with the genetics of heritability and population-level genetics. He obviously leans towards the more gradualist, 'gene as an evolutionary unit' paradigm first popularised by Richard Dawkins in The Selfish Gene. While a useful refresher for readers who might be only slightly familiar with the precepts of natural selection, there is nothing here that is likely to change anyone's mind if they don't already accept the theory of evolution.
Having established the importance of neo-Darwinian evolutionary theory as an underpinning of his argument, Dennett then goes on to explore what he refers to as "inversions of reasoning", one arising directly from Darwin and the other from another, more recent intellectual giant of technology and computer science, Alan Turing.
Darwin's inversion of reasoning is that simple structures and random changes, accumulating over time and under the influence of selective pressures, are able to produce complex structures and behaviours. Dennett calls this a "bubble-up theory of creation" in contrast to the "trickle-down theory" that predated Darwin.
Turing's inversion of reasoning comes from the observation that, with the appropriate set of instructions, a Turing Machine can carry out any arithmetic operation, and thus anything computational, without any sort of "understanding" of arithmetic or the computations being carried out. Dennett terms this "competence without comprehension", and uses this concept throughout the book to help explain how apparently simple structures (well, simple at one level at least) such as neurons, acting without any knowledge or understanding of their specific roles, can result in an organism gaining competence in skills and, eventually, consciousness.
I will admit that I found the first third or so of the book a bit heavy going. In places I was bored by the repetition of material I am already quite familiar with, such as the Modern Synthesis and neo-Darwinian theories of evolution, while in other areas I found myself somewhat lost in the philosophical jargon and having to read and reread passages to gain some inkling of understanding. It is worth the effort though, as when we reach part two of the book things really start to pick up steam.
I can haz sentience naow?
If you wanted to boil down all 400+ pages of the densely packed argument Dennett puts forth into some central postulates, the foundations on which he argues that self aware minds were created (in us at least), these could well be:
- At some point, through processes of Darwinian natural selection in response to some environmental change or stressor human brains evolved to allow:
- The development of a persistent culture, through either oral tradition or symbolic transfer of information, leading to a condition where:
- Thoughts, beliefs, ways of doing things, even words themselves could be transferred from individual to individual and thus could themselves be subject to natural selection, resulting in a cognitive arms race that resulted in:
- Cultural evolution, and the development of both global comprehension of the individual's inner state, and the ability to step outside of the random, highly iterative processes of Darwinian evolution and adopt a more top-down, intelligent approach to problems and a self-awareness that we would call a mind.
Those thoughts, beliefs, ways of doing things, and even words themselves, Dennett refers to as 'memes'. But these are not just the memes of the modern internet, instead he uses the term in much the same way as Richard Dawkins did when he first coined it in The Selfish Gene, memes as units of transmissible culture; essentially, contagious ideas. The memes of the internet, lol-catting their way across our screens, are just a subset of the greater world of memes. Even the word meme is a meme, illustrating Dennett's point about words-as-memes very nicely. We say that a meme, a phrase, a video or saying has "gone viral", and even the term "gone viral" illustrates this idea of contagious ideas and how they can multiply. Very few virologists would say that this year's flu has "gone viral". There is no one way in which viral diseases spread or replicate, viral is an adjective, not a verb. All of this not withstanding, the [neoDawkinsian] meme that can encapsulates the belief that "going viral" means a phenomenon of rapid spread from individual to individual of a contagious idea is obviously a good fit for the modern mind, and is fit (in the Darwinian sense) to survive and reproduce in modern culture.
No Noam, Just No.
Books like this are an important way in which philosophers communicate, and so a lot of time and space is spent both in building the author's argument, and directly addressing other authors and thinkers in the field. Bacteria to Bach and Back certainly holds to this tradition, and in some areas Dennett spends a lot of time, words and thought on answering critics of his earlier work in this area, or revisiting his own criticism of other authors.
One author and thinker who receives quite a lot of criticism from Dennett is the linguist and cognitive scientist Noam Chomsky. Essentially, Dennett derides Chomsky's proposed Language Acquisition Device, a hypothetical "module in the mind" that would be responsible for the acquisition of language skills in children. Dennett views this as a sort of high level handwavium, putting off understanding of the mechanisms of language acquisition to some ineffable, non-locatable and mysterious "thing". Given Dennett's overarching theme that consciousness and the mind are based on the development of words-as-memes and subsequent cultural evolution that is all built upon the relatively simple processes of Darwinian evolution, it is easy to understand his dislike for this position, and he spends several forcefully argued pages, in my mind at least, successfully arguing against Chomsky's position.
As a reader who is only passingly familiar with the larger figures and controversies in this area, at times the careful dissection of other author's points and counterpoints did drag a little, however this is an unavoidable consequence of the book's attempt to act as both a popular science, readable and accessible summary of Dennett's work and a strident, well rehearsed response to his own critics and professional peers.
Singularity, Slavery, or Symbiosis?
As the book and his long and winding argument come to an end, Dennett indulges in some speculation about the future of minds, or "intelligent agents", both human and artificial. After describing in some detail the circumstances via which he believes that self-awareness, sentience and minds as we understand them have come about, and providing several lines of quite pointed criticism of the "top down" intelligent design approach as it applies to the development of artificial intelligence systems, in the closing chapters his argument pivots to the problems and pitfalls that may attend the development and implementation of more and more capable intelligent agents.
In a large part, Dennett's concerns here revolve around the tendency of humans to attribute knowledge, understanding and intent to inanimate systems if they do "anything that impresses them as at all clever". For this tendency he uses a term he himself coined, the "intentional stance", defined as "the strategy of interpreting the behaviour of an entity [...] by treating it as if it were a rational agent". Given human tendencies to adopt this intentional stance, Dennett worries that we will (his emphasis) "over-estimate the comprehension of our latest thinking tools, prematurely ceding authority to them far beyond their competence".
This concern leads to some remarkably robust suggestions, some of which border on hysterical. We should "expose and ridicule all gratuitous anthropomorphisms in systems", so no Siri or Alexa for you. Even more extraordinary, "systems that deliberately conceal their shortcuts and gaps of incompetence should be deemed fraudulent, and their creators should go to jail for [...] creating or using an artificial intelligence that impersonates a human being".
Dennett obviously does not feel that the "top-down" approach to AI research, resulting in expert systems such as IBM's Watson and brute force game playing systems such as Deep Blue and AlphaGo, is the correct direction to develop "true AI", systems and machines that are self-aware and "truly conscious". In his words, top-down AI research has produced "a hyper-competent, but deeply unbiological marvel. This is not because they are made of the wrong kind of stuff but because the stuff they are made of is organised into the wrong sort of hierarchies...".
One area I felt was lacking in this final speculative section of the book was consideration of what, if any, responsibility we may carry as creators of artificial conscious systems, assuming that the methods suggested by Dennett's argument are successful. Having earlier described the modern expert systems produced by GOFAI as "bedridden aspirants to genius", architectures that are "brittle [...] vulnerable [...] and utterly dependent on their handlers", I get the impression that he has given some thought to the moral perspectives of creating new minds. This subject, however, is not picked up further. Instead, he makes a case for the development of "cognitive prostheses [...] designed to be parasitic, to be tools, not collaborators." I can't help but feel that this sells short the potential value that could come to humans from the realisation of self-aware, artificial intelligences that can experience the full richness of a complete and whole mind.