Posted by zompist
https://zompist.wordpress.com/2025/08/09/from-bacteria-to-bach-and-back/
http://zompist.wordpress.com/?p=14758
I’ve just read Daniel Dennett’s last philosophy book, From Bacteria to Bach and Back: The Evolution of Minds (2017). My overall reaction is that it’s a step down. If you’ve never read him, instead try Consciousness Explained (CE, 1992) or Elbow Room: The Varieties of Free Will Worth Wanting (1984).
Bacteria to Bach is long and ambitious, yet breezy and self-indulgent. He was 75 when he wrote it, and I feel like it shows. It’s a summarization of his lifelong views, and it often feels that he is tired of making these points to uncomprehending philistines– can’t they just admit they’re wrong? Unfortunately this attitude doesn’t produce the slow but scrupulously argued framework of CE, but long repetitive rants that assert rather than demonstrate his ideas.
Now, many of his ideas I completely accept:
- The mind is ultimately a machine: no immaterial substances are needed to explain it.
- This does not make it either deterministic or valueless. Our emotions, thoughts, values, and rational proofs are all still there.
- Minds ultimately derive from evolution.
- Evolution itself is a mindless process that mimics intelligent design. Because success is rewarded with reproduction, it is not entirely random, but can both innovate, and spread innovations.
- There is nothing wrong with speaking of biological features as having utility, and indeed design. There’s just no designer.
- Humans, at least, have a new quasi-Darwinian process of cultural evolution, which allows much faster innovation and spread of ideas, this one not tied to inheritance. You can’t grab another animal’s genes, but you can steal another culture’s ideas.
- Humans also have real intelligence– in effect we have an internal computer which lets us imagine and reason, a third way of producing good ideas (and some bad).
- Looking at cognitive abilities, we should always consider them as a continuum from mindlessness to full consciousness– not least because this is the only way we can understand how minds evolved. Searle-style insistence that understanding must be all-or-nothing and human-level is a mistake.
- The “Cartesian Theater”– the screen in the brain watched by a homunculus– is simply not there. There is no central control room in the brain, no organelle which we can label “consciousness.”
- Both introspection and perception can be fooled.
- Our beliefs or claims about our own consciousness must be taken with a grain of salt. A good example is syntactic knowledge: we follow grammatical rules that our brain knows, but which are not directly available to consciousness. (I covered this in the Syntax Construction Kit.) When we say we know a language, it’s true that our brain knows it, but “we“, our consciousness, are only part of our brain.
So, what don’t I like?
First, he has a pair of beliefs that seem to be both dubious and contradictory. One is that only humans have full consciousness: even bears and chimps, to say nothing of bees and bats and dogs and grey parrots and dolphins, may be without real comprehension, there may be no one at home there. The other is that our own consciousness is a user illusion— we do not have special insight into how our minds work, there is no such thing as qualia, and much of our apparent competence is not actually “ours”. It bubbles up from the brain and we take credit for it.
I don’t know why he wants this division between humans and other animals— it’s profoundly out of tune with the last century of biology. If you read Konrad Lorenz you’ll see the deep similarities between human and animal behavior… even down to the level of fish. He actually makes his central task harder– explaining consciousness– if he can’t grant it to dogs and chimps.
He wants to tie human abilities to language… OK, but when a grey parrot can learn 200 words, it gets a lot harder to draw this bright line. His idea is also strangely oblivious to his own repeated and correct point that aspects of human cognition must derive from something simpler.
His chapter on qualia has a bullying tone that gets on my nerves, e.g.:
By offering a sketch of the causes of Cartesian gravity, I have tried to help the unpersuaded find a vantage point from which they can diagnose their own failures of imagination and overcome them.
Sometimes a disagreement does derive from a failure of imagination, but you have to show it, not assume it.
What Dennett is asking for is pretty much what philosophy, and some religions, have been asking us for 2500 years: to understand that we can be fooled, that our mental imagery might be fake. Plato made the same point with his cave analogy, Descartes with his malicious demon. That we don’t really have a soul, but are only a place where perceptions swirl around, is longtime Buddhist doctrine; also Humean doctrine.
It’s a point of view, but I’m always reminded of Samuel Johnson’s response to Berkeleyan idealism: kicking a stone, he says “I refute it thus.” I’d read one anti-qualia passage after another, then look around my bedroom at the exquisite picture of the world presented by my personal brain, and start to wonder, doesn’t Dennett have that? Or: Does he think ChatGPT has that?
Yes, we have dreams, hallucinations, optional illusions, mistakes in perception. But the common-sense perspective that these errors don’t matter much is not disproved. CE actually started with a very convincing argument that the “brains in vats” scientists have an immensely difficult task in front of them. One can recognize that the brain is creating our perceptions rather than just passing them along, but the amazing thing is not that our perceptions can be fooled, but that they’re as good as they are.
See also Anil Seth‘s book on consciousness, which offers the intriguing idea that perception is far more top-down than we might expect. That is, our perceptions are a creation of the brain, which after all is imprisoned in a dark bony cavity with only nerves coming in. We see things, but not with light: there’s no light in the brain. Rather than simply building up a picture of the world from sense data, the brain is creating that picture, then testing and confirming it with sense data.
Dennett reproduces an optical illusion: a color-reversed American flag. If you stare at it for 30 seconds, then look at a blank page, you will see a “properly colored” illusory American flag. It’s an effect easily explained by neural anatomy: nerves exposed to the same stimulus get tired: stare at the green stripe long enough and it looks less green. When you stare at the white page afterward, that part of your visual field is temporarily discounting green, and what that looks like to us is red.
Dennett asks, with the air of producing a gotcha, if the red stripe we see is really red. But it’s a meaningless question. Of course it’s an illusion, there is no stripe in the world, and no redness. Is there a “red stripe” in the brain? There’s a representation of one, yes– just as there was a representation of the “real” green stripe. Why is the difference important? He could go on to show, as C.L. Hardin does, that the stripe on the page need not be green either. (In a printed book, green is produced by a mix of yellow and blue dots.) The “green” is in our eyes or brain. And, well, so what? These things and these illusions are interesting, fun to think about, but don’t actually disprove qualia… they are qualia.
Back in CE, Dennett had some excellent critiques of the Cartesian Theater. E.g. he points out that we literally cannot see in the region of the visual field interrupted by where the optic nerve pushes into the retina. There’s a quite large gap there, a gap we do not perceive. He argues convincingly that it’s wrong to say that the brain fills in that gap. The brain only has to answer questions that it asks, so to speak. There is no part of the brain that asks or needs to be told what’s in that gap.
Part of the answer to this is saccades: the eye is constantly moving, so what’s on the retina is not a fixed image anyway. The brain turns that ever-changing kaleidoscope into what looks to us like a fixed image. But it’s not correct to say that our perceived image is high-resolution. Rather, the brain uses a trick: when it needs to (e.g. for reading), the eye gets a lot of detail, by aiming the high-res part of the retina, the fovea. Because whatever we’re looking at is hi-res, we are not aware that the rest of the visual field is lo-res. We have to be very careful not to treat as facts things we merely suppose about our qualia. That is, just because we’re not aware that our perception is lo-res except for a moving hi-res spot, doesn’t mean that our perception is “really” hi-res.
But the later Dennett seems to have mislaid the plot, denying things that don’t need to be denied. Surely it’s not satisfying to explain bits of consciousness (qualia) by just writing them out of existence. If a mental patient said some of the things Dennett does, they’d be hospitalized. (There are mental conditions where a person thinks they’re dead, or blind, or their legs don’t belong to them.)
The other bit I don’t care for is his extended discussion of memes. He wants to use these to explain cultural evolution, but he never succeeds in showing why memes, in particular, help out. I think he’s seduced by Dawkins’ original parallel to genes, which he famously called selfish: genes can be reified as things that have goals of their own, using animals (and plants) to reproduce themselves! (Biologists way over-emphasized genes in the 1990s.) Likewise memes can be seen as idea-complexes that colonize our minds for their own benefit, that benefit being defined in terms of mutation, and competition to reproduce.
The problem is, Dennett does not actually show that this notion explains anything particularly well, besides what “memes” mean in popular discourse: viral jokes. You can get a frisson of contrarian thrill by picturing ideologies and religions as memes, pointing out that they usually contain antipatterns that facilitate spreading the idea and discouraging abandoning the “faith”. Fine, I included a whole chapter on “stickiness” in my Religion Construction Kit: it’s a useful question to ask not just what a belief system says, but what’s attractive about it and how it protects itself. But memes as an idea… memes as a meme… are strangely underpowered. They seem mostly to be a way of disparaging an idea we don’t like. I don’t think Dennett ever calls evolution itself, or the scientific method, or Bach’s piano music, memes. You don’t need the catchy name to get across the idea that useful ideas spread, or even to point out that “useful” may be a loaded term here.
He tells us that words are memes, in a way that suggests a gotcha that never comes. All in all he talks about language in a way that belongs more to the 19th century than the 21st. It’s profoundly un-linguistic to talk about languages or words competing with each other, and the “better” ones winning. People and nations compete with each other, but neither languages nor words do. Does it mean something that we say “dog” and the Spanish say “perro” and the Japanese say “inu”? No, it really doesn’t: it’s effectively random. Sound and symbol are (usually, not always) separate; none of these words are better and none of them compete.
Now, Spanish replaced Latin “canis” with “perro”. Aha, evolution by natural selection! Only, no. Meanings change, words are borrowed, but “perro” is not better in any way, it did not win the meme wars. Language change is largely value-neutral. Memetics is the cryptocurrency of culture studies, a solution in search of a problem.
If you insist on thinking about the origin of language— generally a futile, self-deluding pursuit– Dennett offers some other, better ideas. E.g. it’s worth recognizing that early language must have been impoverished yet still had to be useful, for either genetic or cultural evolution to boost it. It would not have started with a mastery of 10,000 words… probably not even 1000. His idea seems to be that words started as viral bits of repetition: it was faddish or fun to reproduce these eructations. I don’t see that as compelling, but as a conlanger I do want to ask: what’s the minimal near-language that’s useful? It has to be far far simpler and easier than pidgins, and probably within spitting distance of primate calls and gestures.
Dennett does have other useful insights. E.g. he quotes Emerson Pugh: “If the brain were so simple we could understand it, we would be so simple we couldn’t.” So, maybe the brain is too complex for a human to understand. But he goes on to ask: is it too complex for a group of people to understand? It’s an excellent question. A lot of pop history of science is about lone geniuses, but most advances today are made by small teams… or large teams. Maybe consciousness will turn out to be something that, in effect, requires a whole university to understand.
His central insight from CE is also worth retaining. There he posited that language allowed people to ask each other questions. If this became automatic enough, they could ask themselves questions when they were alone, at first out loud. We can still do this, and it works! It focuses the brain on the question and sometimes an answer pops out. We can imagine the process improving by both genetic and cultural evolution until we get our modern internal stream of consciousness.
The last chapter is on AI, long one of his preoccupations. It’s interesting to see his reaction to early deep learning. Kind of unfortunately, he was writing about 5 years before ChatGPT came out. So he is simultaneously a bit breathless about how great the AI of 2017 was, and skeptical about how good it could get. But he does, to his credit, recognize the problems of overinterpretation:
The real danger, I think, is not that machines more intelligent than we are will usurp us as captains of our destinies, but that we will over-estimate the comprehension of our latest thinking tools, prematurely ceding authority to them far beyond their competence.”
He suggests that AIs, like medicine, should come with long lists of possible shortcomings, and that “systems that deliberately conceal their shortcuts and gaps of incompetence should be deemed fraudulent, and their creators should go to jail”. Too bad no one listened to that.
https://zompist.wordpress.com/2025/08/09/from-bacteria-to-bach-and-back/
http://zompist.wordpress.com/?p=14758
![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
cosmolinguist at 09:28pm on 10/08/2025
okay![[syndicated profile]](https://www.dreamwidth.org/img/silk/identity/feed.png){kind=small}
content