CHAPTER 7: THINKING
QUESTIONS, CRITICISMS, & SOLUTIONS

1. What algorithms control what we focus on?
   • Perhaps we develop subroutines for each of our hobbies and interests that alter our focus when they detect something interesting.
2. Prof. Minsky points out that, as a last resort, you can make a plan by exhaustively searching through all possible plans. Might it be better to randomly generate plans, and keep the best ones?
   • Yes, but not if you're just shuffling the list of all possible plans before looking at them.
3. What methods do minds use for detecting and avoiding certain thought processes (Ways Not To Think)?
   • e.g. learning from mistakes, following a result to a bad conclusion, installing a new Critic, generalizing from a bad experience.
4. How do you prevent subgoals from monopolizing your mind and making you forget what you were originally doing?
   • Sometimes you do forget, or you pursue a short-term goal that violates your long-term goals. But some strategies might be: if you often try to solve the same goal, store the solution process in a script which has no divisions into goals and subgoals; or, maintain a special stack of recent mental states.
5. I think that effectively using what you already know is at least as challenging as inventing a new way to think— they might even be very similar sorts of processes.
   
   What methods do we use to effectively apply what we already know, and which of those methods are also applicable when we're learning something new?
6. How do we develop such complicated ways to think? And how can researchers empirically investigate this development, if so much of our thinking is inaccessible to our own introspection?
   • Maybe “reinforcement” encourages us to make progress with low-level ways to think, and “teaching by example” helps us learn our higher-level ways to think — but psychological terms like these don't explain how we construct the behaviors we're rewarded for.
7. How do reflexes and muscle memory relate to the Critic-Selector model?
   • Perhaps the Critics originating in the lowest levels of the six-layer model have the quickest response time.
   • Also, more generally, it would be interesting to see what parts of our mind-software are implemented in hardware beyond the brain— for example, in the spinal cord and muscles.
8. How do our Critics change over time— and what processes would change them?
9. How do we manage with such imprecise, non-predictive models of other people (and of ourselves)?
10. How do we keep track of what we know?
    • For example, what models of self do we use, and what kind of common sense information to we have, and what storage/retrieval methods do we use)
11. How can we describe the phenomenon of thinking outside the box in terms of the theories proposed in this chapter?
12. How do we restore our previous mental states so rapidly?
    • Our short-term memory is probably optimized for rapid re-enstatement of previous mental states, but is probably very limited in space. (If we put too much on the stack, we'll lose some of the things on the bottom.)
13. What ways to think do we use when we're sleeping? Can we characterize different phases of sleep by the ways to think we use during them? Which ways to think are exclusive to sleeping? Exclusive to waking? Common to both?
    • It's clear that much of our visual processing routines are offline when we sleep, because our eyes are closed. I suspect we use our sleeping time primarily to re-organize our minds while they're less occupied.
14. Many people report having great ideas in their sleep. Is sleep conducive to the creation of good ideas?
    • I suspect that sleep turns off many of our usual Critics, which can have both good and bad consequences:
      When some Critics are off, we might examine solutions that we would otherwise rashly reject. On the other hand, we might overlook the flaws in a solution and believe that it's impossibly perfect. Then, when we wake up, we're dismayed that we can't remember the details well enough to reconstruct it—every solution we think of when we're awake has an obvious flaw in it. As a result, we begin to believe that our best ideas are produced when we're asleep, because they seem to have many fewer flaws.
15. We have some ability to direct our ways to think, but most of it happens automatically. How do the automatic processes (e.g. habits) interact with our conscious overrides?
    
    (Thinking has the same quality as breathing: it happens automatically, but we can partially control it when we think about it.)
16. How can we learn to override our instictive responses? (That is, what mechanisms are responsible for our change in behavior from reacting instinctively to everything, to reacting more deliberately?)
    • As suggested in Chapter 2, I think that as our higher levels of thought mature, they become able to control and modulate our reactions more smoothly— but I would still like to know how that works.
17. If negative thinking has more influence than positive thinking, does this mean that critical words are more formative of our behavior than positive words?
18. How can we design a machine that “takes a break and subconsciously keeps thinking about a problem”?
    • Perhaps inaccessible subconscious routines are an evolutionary hack that machines don't need. If that's the case, the machine could just restrict the amount of resources it spends solving the problem and reflecting on its progress.
19. How do we stop ourselves from overgeneralizing or oversimplifying a problem?
    • Perhaps we have general-purpose routines for detecting lack of progress in any endeavor, or we have special-purpose routines for recognizing previous overgeneralizations.
20. What determines how people react to criticism?
    • I suspect that the more Ways to Think you have in a domain, the better you react to criticism in that domain, because it's less likely that any particular criticism will make you become stuck and frustrated.
    • Maybe that's why it can be so frustrating to learn a new skill, unless you have a teacher who can show you how to apply what you already know to the new domain.
21. How do mental disorders affect Critics? Are mental disorders a kind of Way to Think?
    • While some mental disorders may be the result of faulty hardware, I think that most take place in software. I suspect our society censures some Ways to Think that it considers harmful or maladaptive, while praising others that it considers beneficial. (e.g. we praise geniuses but condemn lazy people and people with schizophrenia) All are Ways to Think, but we select for the ones our society likes best.
22. What is the function of psychiatric therapy, in terms of our more sophisticated model of the mind?
    • Perhaps therapy helps us alter parts of our brains that we couldn't ordinarily “reach”, by learning new ways to think and new ways to model ourselves.
23. Criticism: I don't think that minds can have software diseases. In humans, every problem eventually becomes a physical hardware problem— sadness causes chemical imbalances which result in physical disorder. We should abandon the hardware/software analogy in humans, and seek a different representation.
    • Editor's remark: Just like a you can run a software virus on brand-new hardware, wrecking the operating system while leaving the hardware unharmed, you can similarly have undesirable thoughts in a brain that's working exactly as it was built to.
    • Every mental problem (e.g. being confused, being depressed) corresponds with a difference you can detect in the brain. However, not every mental problem corresponds with a problem in the brain (i.e., malfunctioning neurons).
24. How do children develop ways of thinking?
    • I think that children have sophisticated programs for learning and organizing new things — perhaps more complicated than most adults have, because adults require those skills less.
    • The thesis that children have more sophisticated learning programs than adults is supported by the fact that some skills can only be learned in childhood.
25. How does confabulation happen?
26. Criticism: Prof. Minsky suggests that our model of a single self prevents us from noticing that we switch between vastly different ways to think. I would argue that we do often regard ourselves as having multiple parts, and we view ourselves in the third person — so perhaps the single-self model is merely highly distracting.
27. How do individual children learn to use obstacle-overcoming strategies like changing the subject, self-determination, self-conscious reflection, self-regression, cry for help, and emotional thinking?
    • Processes like these are so sophisticated! It seems impossible that they could be innate evolutionary adaptations, because the infants that possess them have never seen the world or developed a model of self. On the other hand, it also seems impossible that such strategies could be learned through trial and error, which is how children learn. Also, it's remarkable that so many people learn the same strategies.
28. I see how you could hardwire a brain to have certain Critics and selectors, but how could you generate a new Critic, or a new representation?
29. What are effective strategies for making plans?
    • Perhaps we could borrow Poincare's strategies for making discoveries: Incubation generates many potential plans, Revelation recognizes a promising plan, and Verification confirms that the plan is good, and records some of the successes involved in constructing it.
30. How can we make an AI that can develop its own individuality, if our current design strategy is to hard-code the rules, beliefs, and opinions of our programs ourselves? Moreover, how can we make an AI that not only pursues its own interests, but also excels at collaborating with others?
    • We need to identify the most efficient ways to communicate and collaborate, and to design AIs that are skilled at collaborating, and which have collaboration as a high-priority goal.
    • Hopefully, it's a general truth that the most effective strategy for getting what you want is to interact positively with your neighbors. If this is true, then the most intelligent robots will figure this out, and will consequently also be the nicest. (It's better, but also more difficult, to make everyone happy rather than just yourself.)
31. Suggestion: Perhaps the imaginary friends of children serve the same purpose as self-models and imprimer-models in adults:
    
    Just like we internalize value systems (we progress from believing that something is wrong because it's punishable, to believing that it's wrong because it's wrong), we also internalize our models of self and virtue.
32. How do Critics contribute to personalities?
33. Our patterns of thinking seem controlled by the environment — when a deadline looms, I'm more analytical and meticulous; but when I'm on holiday, I'm more open-minded. Why can't I deliberately switch between these different modes?
    • I suppose it's evolutionarily dangerous to be able to do so; still, you can coerce yourself to think a certain way by using the "Professor Challenger" trick of imagining something that makes you think a certain way, or by putting yourself in a situation (e.g. listening to a nostalgic song) which activates some of your K-lines and puts you into the appropriate mental state.
34. Why are people hypocritical in the following sense: they are most outspoken against views that they formerly held, or vices that they (used to) have?
    
    I suspect it's because their self-Critics are highly developed and sensitive, and so they trigger cascades of emotion even when the Critics are activated by other people instead of one's self.
35. What if emotions and thinking are not independent faculties, but instead emotions are just another Way to Think?
36. What is the role of memory with our instinctive, reactive layers?
    • I suppose low-level layers don't activate memories themselves, but perhaps memories activate them. (e.g. a frightening memory can cause fear). Memory is undoubtably relevant in all layers of the mind.
37. F. Scott Fitzgerald once said, “The test of a first-rate intelligence is the ability to hold two opposed ideas in the mind at the same time, and still retain the ability to funciton.” How might we accomplish this feat?
    • Prof. Minsky argues that we can only maintain a few high-level thoughts at once, because they all rely on the same resources or because they exist in specialized newly-evolved hardware. Perhaps we achieve this feat when we are so economical that our carefully-placed K-lines can maintain (and switch between) conflicting records in a single context box.
38. If thinking is not all linguistic, then what other representations do we use?
39. What are effective strategies for learning to be more creative?
    • Perhaps we can methodically investigate the strategies that creative people use — for example reformulation, reasoning by analogy, use of external representations, wishful thinking, and impersonation— and deliberately practicing them ourselves. It would be more convenient if we had finer control on our subconscious, though.
40. Can the method of analogies relate concepts between different disciplines?
    • Editor's note: Yes.
41. How much of our thinking is true parallel, i.e. several different processing happening at once, and how much is timesharing, i.e. actively storing and re-storing mental states? Do both kinds of multi-tasking happen simultaneously?
42. It seems strange that the brain both stores data and processes it, sometimes simultaneously— might the brain have specialized subdivisions for storing data, and for processing it?
43. Criticism: I don't think that our ancestors developed many different ways to think in order to compensate for the variability of the environment.
    
    After all, there are probably only a few possible variations in the environment, and only a few different survival problems— even so, there are still many different kinds of people and talents.
44. Criticism: Prof. Minsky's ideas on using bodies as memory devices seems outlandish; I doubt that our facial expressions can really control our emotions, and I think it's the other way around.
    • What happens when an accident prevents you from, say, clenching your fist or making a certain expression? Surely we have redundant fallback systems so we can still feel emotions even after some injuries— but perhaps the emotions are suppressed, or the brain has to rewire itself to compensate.
    • Editor's note: Facial efference.
45. Criticism: I think that expression-caused emotions are an evolutionary hack; while they may give insight into human behavior, they aren't a necessary feature of generally intelligent robots.
    • On the other hand, we can learn a lot of new Ways to Program by studying our brain's exotic hardware and comparing it to, say, von Neumann architectures. Perhaps facial efference is part of a class of efficient storage+reenforcement+communication mechanisms we haven't yet explored.
46. When you talk about the patient in Damasio's book, I don't understand what you mean when you say that “his inability to make decisions resulted in his lack of emotions”. Shouldn't it be the other way around?
    • I've been told that emotions interfere with decision-making abilities. Surely a dispassionate person would be more decisive?
    • Why would being stuck in a perpetual “indecisive” state of mind also affect your ability to transition into different emotional states of mind?
47. What happens differently in the mind when you give up on a problem, compared to when you set it aside temporarily?
48. How do supervising agents measure time, for example to tell that you've been trying one thing for “too long”? Relatedly, how does the phenomenon of immersion happen, where we become less susceptible to distraction the more we become involved in a task?
    • I suspect that saturating your short-term memory with your current task has the effect of encouraging you to continue to do it. (Although this does not explain why boredom does not become increasingly interesting the longer you experience it.)
    • Maybe as your higher-level agents begin making and executing their plans to further your specific task, they take control of resources that would otherwise be used to notice distractions in the environment. Perhaps the longer you work, the more of a monopoly they acquire — but what opposing force prevents you from becoming a monomaniac? I suppose our built-in Critics (or critical friends and family, anyways) prevent us from doing any one thing for too long.
49. Sometimes, after we find one solution to a problem, we spot a second solution. It seems like there must be a highly complex collection of mental events that causes us to find those solutions in that order. How can we track down the fine-grained mental operations involved?
50. I've notice that I can usually find a second solution to a problem, as long as I found the first solution myself. Otherwise, I find it terribly difficult. Why?
    • In the process of finding the first solution, I activate a lot of resources that presumably help me find the second. I suppose when I'm handed a solution, my high-level agents become fixated on that paradigm, and have trouble activating a bunch of apparently unrelated resources in search of a different solution. Maybe a Critic even shuts them off to prevent them from wasting resources!
    • Relatedly, spoken answers are described in a summarized way that doesn't allow you to see which of your resources you would have needed to use in order to solve the problem. This means that only a few, superficial credit assignments can be made and so little improvement takes place.
51. What mechanisms allow us to focus, and what processes control how "deep" the focus is? What conditions make it easier to return to a task after setting it aside temporarily?
    • I imagine there are Critics which are supposed to prevent me from doing things that are boring or unimportant. This explains why it's harder to return to unimportant tasks, and why diversions that I enjoy are more persistently distracting than, say, environmental noise.
    • But how can we overcome procrastination and other flaws of focus, to help us concentrate on things that we wish we cared more about?
52. Idea: Prof. Minsky suggests that it's difficult to read and write at the same time because both skills rely on the same resources. Maybe the reason why some people can't concentrate with background music is that they developed listening-to-music skills which rely on the same resources as their working skills. (Or perhaps people who can listen to music have better Critics for ignoring music.)
53. How do we develop new context-boxes / realms of thought? Are context-boxes discrete, or do they overlap and meld into each other?
54. How do brains recognize when a new Critic is necessary and, more importantly, how do they create a new one (e.g. to recognize a new class of situations)?
55. Why don't most animals invent new Ways to Think?
    • I think most people would say that it's a hardware deficiency— animals don't have the specialized brain modules that we have. On the other hand, it may be more an issue of software— we are born with better ways of rearranging our brains. If we didn't have the software, our brains would remain animal-like and static.
56. Why don't we remember everything? Suppose we could add auxillary memory to our brains— wouldn't it then be a good idea to remember basically everything, as long as we could prioritize what we knew, and could garbage-collect it from time to time?
57. I am interested in the Wishful Thinking strategy. Might there be more precise variants, other than trying to solve the problem with unlimited resources?
    • For example, you could methodically turn off collections of Critics to reduce the number and kind of constraints on the solution. (Turning off most of them would be like giving yourself infinite resources, or would make you feel like the problem was already solved.)
58. In the Resignation strategy, what is the entity called the “Rest of your Mind”?
    • I suspect it's some part of the unconscious mind (i.e. the parts whose records aren't available to the language or other high-level agents.)
59. When reasoning by analogy, how do we decide on the relevant and irrelevant features? Also, if we find more than one plausible mapping, how do we decide on the right one?
    • I imagine we throw out mappings that don't make sense to us, and prefer mappings that are simpler or more clever.
60. I agree with Prof. Minsky when he says “it ought to be one of our central goals [...] to classify our Ways to Think. However, we don't yet have systematic ways to classify those abilities.”
    
    What progress has been made to organize our different Ways to Think?
    • I think that some strategies can employ others, so you could arrange them that way. For example, Reasoning by analogy could uncover a situation in which Divide and conquer applied.
    • Perhaps you could organize them by how effective they are. But such an organizational scheme would undoubtably vary from person to person, and even for an individual as he learned new ways to think.
    • In any case, I assume the ideas don't form a simple hierarchy— perhaps a heterarchy, with different organizations for different purposes?
    • We may need better ideas about the components of these ways to think, and their interactions, before we have a sufficiently complete beginning.
61. What kinds of goals do we have when we are solving artistic or creative problems? For example, what are you trying to achieve when painting, and how do you know when you've done it?