Marvin Minsky "The Emotion Machine": Chapter 5 "Levels of Mental Activity"

During the HYIP on the digital economy and ***, it's time to pay attention to how to bring “human” into technology and how technology helps to understand and improve and scale “human”. This will be helped by the stern Marvin Minsky, who analyzes with his merciless mind feelings, emotions, pain, love and consciousness.

Chapter 5. Levels of Mental Activity

“We are obviously a unique species due to the ability to create symbols, as well as the ability to control the conditions of our existence using these symbols. Our ability to represent and model reality implies that we can approximately evaluate the different orders (bitness) of being and ... gives us the feeling of managing our own experience. ”
- Heinz Pajels, “The Dreams of Reason”

No man has the strength of an ox, the secrecy of a cat, or the speed of an antelope, but our species surpasses all others in the ability to invent new ways of thinking. We make weapons, clothing, and homes. We are constantly developing new art forms. We are unsurpassed in creating new social agreements, creating complex laws that we initially apply, and then looking for all sorts of ways to evade them.

What allows our minds to generate so many new things and ideas? This chapter will propose a scheme that reduces all our resources to six different levels of processes.
')

Starting from simple instinctual reactions, each successive level is built on the previous level until they embrace processes that involve our highest ideas and personal goals. In order to understand why we need so many levels, let's look again at the example that was introduced in §4-2.

Joan walked halfway along the road, hurrying to the meeting to bring the final report on the work. While she thinks about what to say at the meeting, she hears a sound that makes her turn her head and she sees a fast approaching car. Being hesitant about what to do next - to run forward or backward, but being influenced by the fear of being late, Joan decides to quickly run across the rest of the road. She later recalls her knee wound and reflects on her impulsive act. "If my knee could not stand and I would not cross the road and be killed - what would my friends think about me?"

The first chapter of this chapter will show how each level of the above diagram can explain what was happening in Joan’s mind at that moment. We constantly react to various events "without thinking", as if we are governed only by the rules "If -> That", which were described in chapter §1-4. But such a simple response scheme can explain only a few of the first events that unfolded in the above story, while the remaining events can be explained by various activities that took place at other levels of Joan’s thinking.

Congenital, instinctive reactions: Joan heard a sound and turned her head. All animals have similar instincts that help them survive.
Learned reactions: She saw a car approaching fast. Joan should have known that such conditions should be addressed in a specific way.
Peer Intelligence: To decide what to say at a meeting, she considers several alternatives, and tries to decide which of the alternatives is the best.
Reflective thinking: Joan reflects on what she has done. She reacts not only to recent events of the world, but also to the recent activity of her brain.
Self-reflective thinking: Being in a state of "anxiety by the late arrival" requires her to follow the plans that she has set for herself.
Emotions of self-awareness: Asking her friends to think about her, Joan also asked herself how her actions fit her established ideals.

The second part of this chapter will show how such systems can "represent" things. Whenever you ask, “What happens if,” or express any hope, desire, or fear, you try to suggest a possible development of further events. Whenever you interact with your friends, you assume how your actions can affect them. Whenever you see something, you assume that a given object can do in the future. How do our mental resources understand ideas that do not yet exist, and then use these nonexistent ideas for their own purposeful change and expansion?

§5-1. Instinctive reactions

“... It shows that all laudatory reviews about knowledge, as about a magnificent thing, are embellished and only 1 out of 40 can be real”
- Mark Twain in "Tom Sawyer Abroad."

Joan heard the sound and turned her head.

Although we live in densely populated cities, there are many birds and squirrels around, and sometimes you can see a skunk or even a raccoon. Although many toads and snakes have disappeared in recent years, countless petty creatures still remain.

How are these animals still alive? First, they had to find enough food to feed themselves. Then, they must be able to protect themselves, because other animals are also looking for food. To regulate their temperature, they must build various shelters and nests. They have a desire to reproduce (and if not - then their ancestors have not developed such mechanisms), therefore they need to look for a mate and raise the young. Thus, each species had to develop a certain machinery, which would allow their youngsters to do different actions and things, without any previous knowledge about them.

Thus, this suggests that they should use some of the selected “If -> That” rules, like the ones described below:

• If something touches the skin, then whisk it away.
• If it didn't work, then move your body.
• If the light is too bright, then turn away the face from the light source.

However, only a few of the “If -> That” rules are as simple as the ones described above, because most of the rules of human behavior depend on the mental context in which we are now. For example, a rule like “If you see food, then eat it” will force you to eat all the food you see, whether you are hungry or if you need it. Thus, such “If” should include a reference to some goals, for example, “If you are hungry and you see food ...”. Otherwise, you will sit on every chair that you see, or constantly switch the switch in front of which you are standing.


How does the above relate to emotions and feelings? If you move your hand quickly towards the fly, this fly will quickly fly away, and this action may tempt us to attribute certain feelings to this fly, such as fear. However, we know enough about the nervous system of insects to confidently say that insects cannot support the complex cascades of nervous activity, which we define as emotions.

In any case, a similar model of "stimulus-response" or "conditioned reflex" was quite popular in the early years of psychology. Some researchers have suggested that this theory can describe all human activity. However, there are some problems in this theory that did not allow this.

The first problem is that most rules have exceptions. For example, if you drop an object, it may not fall if something can prevent it from doing so. Your wristwatches will show you time, but not when they are not working. We will be able to solve these problems only if we introduce some exceptions for the above rules - however, in some cases, there will be exceptions for the exceptions themselves.

What happens if the situation follows several different rules If? In this case, we should have a way to choose one of these reactions. One way is to start organizing these rules according to a certain priority queue. Another way is to choose such If which you use most often. Another way is to choose the used If the rule is completely random.

However, when we face the most difficult problems, the simple rules “If -> That” will not work, because we need to determine how the situation will develop in the future and what result we can get in the end. So, now we will briefly talk about the threefold rules that serve to predict each of our actions.


If we have an adequate set of “If -> That -> In this case” rules, then we can answer the question: “What will happen if ...” before we take any action? Then, doing this operation cyclically, we can present a much more complex decision tree. We will soon return to the discussion of this topic, but for now we will discuss exactly how systems can learn the simplest rules "If -> That".

§5-2. Studied Reactions

All animals are born with “instincts,” such as “moving away from a rapidly approaching object.” Such built-in reactions usually serve well as long as the animals remain in an environment similar to the one in which these instincts were acquired. But when the environment changes, these same creatures must work out a completely different way to respond to environmental hazards. For example, when Joan heard an approaching machine, she partially acts on the basis of instincts, but she also reacts on the basis of her knowledge of the nature of this danger. But how does she study and what exactly has she studied? We will return to this issue at the end of this book, because teaching people is incredibly difficult, and here we only briefly mention how learning can take place in some animals. During the 20th century, many well-known psychologists had the following beliefs, as animals learn new rules for “If -> That”:

When an animal encounters a new situation, it tries a random sequence of actions. Then, if any action gives any "reward" this action is "fixed". This causes the animal to use this action in any such situations.

This theory of “learning through reinforcement” can be used to describe the absolute majority of the activities that animals do. In fact, this theory is based on experiments conducted in mice, rats, parrots, dogs, cats and snails. However, this theory does not help explain how people learn to solve complex problems whose solutions require a series of complex actions. Indeed, deciding what to learn after solving a difficult situation can be much more difficult than directly solving a complex problem, and the words “case”, “reward” and “consolidation” do not help us answer the following two difficult questions:

How do successful reactions occur? To solve a complex problem, an intricate sequence of actions is usually required, in which each subsequent step builds on the previous one. A successful assumption can lead to a successful step, but the accidental origin of this step leads to the fact that creating a chain of successful steps would take too much time. We will discuss this issue in more detail below in Search and Planning.

What aspects of recent events need to be memorized? In order for “That” to work well, it is necessary that this rule contain only relevant functions, because it can be brought down by irrelevant functions. (If you learned a new way to tie a knot, then you should not remember that you tied it on a certain day or week). As we will see later in Chapter 8 of resourcefulness, if your description of “That” is too specific, then it will very rarely correspond to real situations. But if your description of “That” is too abstract, it will be suitable for too many situations, and in any case you will not be able to learn a lot from various situations.

For example, suppose you want to create a robot that recognizes the visual image of a human hand. This is quite a difficult task - because we never see the same picture twice, even with the same hand, because each finger can change its position and shape, we can view the hand from different angles of view, and each part can capture different amount of light. This means that we will need to use trillions of “If -> That” rules until we can start using some tricks that will help to highlight only the most important functions, or, as we will see in §6-2, we can formulate high-level descriptions, for example: “an object in the form of a palm with fingers attached to it”.

Of course, many things that we do are based on the reaction to external events using the rules "If -> That". However, along with these low-level reactions, we always draw up new plans and think about what we have done in the past - and it is these inner mental activities that are responsible for the manifestation of our unique abilities.

For example, when Joan reacted to a moving car, her reaction was half instinctive and half learned. However, she could not learn “from her own experience” that cars traveling at high speed are dangerous - because if she studied this fact with the help of trial and error, she would probably already be dead. Being trained on the basis of "reinforcement" of successful actions is a really bad way to learn survival. Instead, she “understood it” on her own, or she heard this knowledge from someone else, and both had to use higher forms of mental activity to gain knowledge. So let us now turn to what we call “thinking,” that is, to the technique that we use to respond not only to the things of the world, but also to the things that are happening in our heads.

---

For the translation, thanks to Stanislav Sukhanitsky. Who wants to help with the translation - write in a personal or mail magisterludi2016@yandex.ru

about the author

Marvin Lee Minsky (Eng. Marvin Lee Minsky; August 9, 1927 - January 24, 2016) - American scientist in the field of artificial intelligence, co-founder of the Laboratory of artificial intelligence at the Massachusetts Institute of Technology. [ Wikipedia ]

Interesting Facts:

• Minsky was a friend of critic Harold Bloom from Yale University (Yale University), who spoke of him as “sinister Marvin Minsky”.
• Isaac Asimov described Minsky as one of two people who are smarter than himself; the second, in his opinion, was Karl Sagan.
• Marvin is a robot with artificial intelligence from the cycle of Douglas Adams novels Hitchhiker's Guide to the Galaxy and Hitchhiker's Guide to the Galaxy (film).
• Minsky has a contract to freeze his brain after death in order to be “resurrected” in the future.
• In honor of Minsk named the dog of the protagonist in the movie Tron: Legacy. [ Wikipedia ]

---