« PREV
NEXT »
« Return to the main page
CHAPTER 7: THINKING
QUESTIONS, CRITICISMS, & SOLUTIONS
- 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.
-
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.
-
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.
-
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.
-
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?
-
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.
- 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.
- How do our Critics change over time— and what processes
would change them?
- How do we manage with such imprecise, non-predictive models of
other people (and of ourselves)?
- 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)
- How can we describe the phenomenon of thinking outside the
box in terms of the theories proposed in this chapter?
- 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.)
- 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.
- 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.
- 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.)
-
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.
-
If negative thinking has more influence than positive thinking, does
this mean that critical words are more formative of our behavior than
positive words?
- 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.
-
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.
-
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.
-
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.
-
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.
- 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).
- 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.
- How does confabulation happen?
- 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.
-
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.
-
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?
- 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.
- 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.)
- 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.
- How do Critics contribute to personalities?
- 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.
- 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.
-
What if emotions and thinking are not independent faculties, but
instead emotions are just another Way to Think?
- 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.
-
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.
- If thinking is not all linguistic, then what other
representations do we use?
- 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.
-
Can the method of analogies relate concepts between different
disciplines?
- 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?
- 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?
- 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.
- 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.
-
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.
- 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?
-
What happens differently in the mind when you give up on a problem,
compared to when you set it aside temporarily?
- 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.
-
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?
-
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.
-
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?
-
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.)
-
How do we develop new context-boxes / realms of thought? Are context-boxes discrete, or do they
overlap and meld into each other?
-
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)?
-
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.
- 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?
- 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.)
- 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.)
- 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.
- 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.
- 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?