Notes from the book “The Mind is Flat”

I finished reading the book “The Mind is Flat” by Nick Chater and here are my notes.

1/ The core idea of the book is that our thoughts are not reflections of a deeper self. Rather they’re real-time fabrications of the mind trying to be consistent with previous fabrications.

2/ In this sense, the book advocates quite successfully, there’s no “unconscious”. There are no deep motivations guiding your conscious self. The conscious, real-time thoughts are all you’ve got.

3/ To understand this better, consider how fictitious stories are written. The novelist has a basic idea of the story but then has to invent details as he proceeds in his writing.

4/ Initially, there’s a ton of leeway in details. Is the protagonist tall or short? What does the house where the protagonist lives look like? But once these details are penned, they constraint the rest of the details in the story.

5/ But notice that the details in the story can never be as rich as the real world. If the novelist failed to mention the color of the protagonist’s tie, it simply doesn’t exist and we’d never know this “fact”.

6/ No matter how hard the novelist tries, the story will always be sparse as compared to the richness of the actual world. (In the actual world, ties always have colors and, unlike in a story, we can always find it out).

7/ Note also that, unlike the real world, details in the story can (and often are) inconsistent. Even the most careful novelist like J. R. R. Tolkien (who invented an entire language for his books) ends up writing story details that conflict with each other.

8/ In The Mind is Flat, the author describes our mind to be exactly the same as a novelist.

That is, our thoughts are:

  • Generated one at a time
  • Have no “hidden” content beyond what’s conscious to us
  • Often, but not always, consistent with previous thoughts

9/ The way I look at it we can either assume that a deep, core mind exists that contains our true self and that true self “bubbles” up to the conscious mind in terms of our beliefs, thoughts, perception and desires.

10/ Or we can assume no core exists and our beliefs, thoughts, perception, and desires are generated “on the fly” as and when we need them. This is the right interpretation of our minds.

11/ The sense of wholeness of our conscious self comes from the fact that newly generated thoughts are consistent with previously generated thoughts, so it appears like our conscious mind accessing the whole rather than one fragment at a time.

12/ This illusion of the sense of wholeness despite the mind’s capability to grasp only 1 thing at a time is illustrated by a series of experiments that show how we read only one or few words at a time despite our feeling of looking at the entire page of text simultaneously.

13/ In a clever experiment, scientists tracked the gaze (fovea center) of the subject and as the subject was reading, they randomized the characters at all other places outside a narrow window around the gaze. Surprisingly, subjects don’t detect any difference. For them, the text appears unperturbed.

14/ The consequence of the book’s claim is that various ideas we’ve taken for granted about our mind are false or an illusion. For example, it’s generally said that the brain keeps on working on a hard problem while we’re sleeping or busy with some other problem.

15/ But this can’t be true if what we know about the brain’s architecture is right. Our brain consists of smaller networks/loops of activity that compete to gain access as the current conscious thought.

16/ Only one of the networks can be active in the conscious mind at a time (recall we read one word at a time), and hence there’s no capacity of the brain to hold multiple simultaneous thoughts. (This is the origin of cognitive dissonance).

17/ So our brain cannot possibly be working on a hard problem “in the background” while we’re doing something else (as that something else is the dominant thought and we process only one thought at a time).

18/ What really is happening instead is that brain switches between different loops/networks from time to time. So, the experience of getting unstuck from a difficult problem is simply the fact that when the brain gets back to the problem, it starts searching for a solution from a different place than previously.

19/ So, really brain is solving the problem in real-time, but we have an impression that the brain must have solved a difficult problem in the background and it surfaced it to us when it got solved.

20/ What I loved about the book was its suggested equivalence between writing a story and having beliefs about oneself. For example, if we see a stranger on the road being sad, we can come up with reasons for her sadness. Perhaps she lost a loved one, or perhaps she didn’t get the job she was hoping for.

21/ When we, of course, ask the stranger directly why is she sad and she may say she is sad because her favorite restaurant closed. But instead of assuming she has access to some “mental depth” about her emotions, she is doing the same as what we were doing: fabricating reasons based on available evidence.

22/ The stranger’s mind may have access to additional facts (such as seeing the restaurant as closed). But if knew the same facts as she did (her favorite restaurant and the fact that it was closed), you would have converged at the same conclusions she did.

And, obviously, you don’t have access to her “unconscious”. So, why assume she has access to it?

23/ The upshot of the book is that when we ask someone how they’re doing, their response isn’t revealing a deep emotion but rather suggesting the best explanation their mind could imagine that accounted for their recent memory traces, bodily signals, and sensory perceptions.

24/ The book also nicely explains why we shouldn’t be surprised when people say something, but do something else. If there’s no deep inner-self existing, just like inconsistencies in a story, such incoherence in our beliefs and actions is entirely expected.

25/ The best part of having no deep self? We can reinvent ourselves whenever we want. Who we are is not set in stone, we can literally imagine ourselves to be a different person, one thought and action at a time.

26/ In the rest of the thread that follows 👇, I’m simply going through the notes I made at the margins of the book and typing them up here.

They’re somewhat randomly ordered, but they will reinforce the core ideas in the book.

27/ We’re not a reflection of some rich inner-world. We’re simply coherent imagination generators.

28/ Our explanations are sparse (underspecified) and incoherent.

29/ The symbolic AI approaches failed because when we tried modeling computers based on what people knew, we ran into the problem that people gave out inconsistent explanations & often they themselves didn’t know how they did something so they fabricated the process.

E.g. you can’t ask someone how they recognize a cat. That just happens by itself. They have no idea.
Similarly, you can’t ask a chess grandmaster how they play so well. They’ll fabricate something.

30/ Data-driven ML/AI approaches work because why learn from people via their explanations when you can learn directly from the world (via data).

31/ Unlike popular belief, there are no multiple selves hidden inside us. This illusion is due to our desires and actions being incoherent, giving an appearance of multiple selves.

32/ When thinking about something, we don’t have a perfectly consistent model of the entire thing in our heads. Rather we have a set of predictive models about different aspects of the problem that collectively give a sense of understanding.

33/ The mind interprets electric signals coming from the body, generated within and from the outside world as meaningful. The signals themselves contain no meaning. Where does this meaning come from? Ultimately, what we find meaningful is what is evolutionarily useful. (What gave/gives us an advantage)

34/ In that sense, our mind is a pattern matcher. It tries to make sense of electric signals by continuously searching for relevant “meaning-patterns” to match against it, and also by creating and storing into memory new composite “meaning-patterns” (based on previous patterns).

35/ If you’ve not seen this before, look at the random blob of white-and-black below. After a while, an object should pop into your conscious mind and you won’t be able to “unsee” it. Your mind has successfully locked onto a pattern.

36/ The reason we don’t detect or care about inconsistencies in our thoughts, actions and beliefs is because most such inconsistencies aren’t evolutionarily meaningful (some might be actually harmful).

37/ What we’re aware of is what’s immediately available for conscious perception. So even though we gaze at a very narrow region in the field of view (our point of fixation is tiny), we feel like seeing the entire scene because if we want to examine something, our eye can immediately access it.

38/ It’s possible to bias people towards a choice that they’ll feel like making freely. This makes me worried: how many of our free choices are serving other people’s interests?

39/ We choose options depending on how easy is it to come up with explanations justifying them (and not necessarily how true they are). This is a real-world Ockham’s razor.

40/ User surveys / interviews are unreliable because slight variations to the “same” question can get us wildly different answers.

41/ Because neurons in our brain are so highly interconnected with each other, multitasking is an exception rather than the rule. We cannot compartmentalize tasks easily. Each activity on a task influences another activity. Our brain doesn’t contain neatly separable neural processes. They’re all interconnected.

42/ Subcortical regions in the brain (thalamus) set the stage for consciousness while cortex fills in with the details. This is why loss of different cortical regions impacts what’s in the consciousness but not the capacity for consciousness. While anesthetics that act on subcortical regions switch off consciousness itself.

43/ We’re trapped in our sensory world, even for abstract thinking. Try an abstract thought (like the number 3) – it’s always in one of the sensory perception modes, either the visual numeral 3 or 3 dots or so on.

44/ Activities such as walking, chewing, breathing are driven by autonomous nervous systems that share relatively fewer connections with the rest of the brain and hence we’re able to perform them simultaneously while holding other conscious thoughts (which require coordinated activity of the entire interconnected brain).

45/ Specialized networks for well-practiced activities like driving can develop because “neurons that fire together, wire together“. Hence, practice “hard-wires” certain networks that are able to function “independently” of the rest of the brain.

46/ Flashes of insight / “aha” moments are suspicions about finding good starting points for a creative process, not a crystallization of the entire creative output (which happens on the go during the creative process).

47/ The same incoming sensory data can be generated from an infinite number of different objects. This inverse optics problem is solved by the brain by latching an interpretation (or cycling through a few) that are evolutionarily meaningful.

Multiple inclined planes make the same image on the retina (via here)

48/ I suspect the entire 2D image in the retina -> 3D perception in the brain can’t be completely hardwired through our genetics. Do our genes have enough information capacity to specify all neural connections required to solve such a hard problem?

49/ Brain infers the current state of the world (including the self) by integrating sensory data, memories, and body signals into the best-fitting pattern it is able to find.

50/ Expertise in a field is not built through “higher” mental power. We all share more or less similar brains. Instead, expertise is built through layers and layers of capacity to recognize (and act upon) meaningful patterns in the respective field.

51/ What makes us unique is that no two people have the exact same experiences.

52/ Imaginative leaps are the acts of connecting two different patterns. That is the whole basis of intelligence.

53/ Thinking is searching through explanations space, one imaginative leap at a time.

Have any comments? Email me. And please share the essay if you liked it: