Will robots inherit the earth?

They say they have been waiting for the promised three years. Here I am in the comments on the article vmb promised to translate the chapter from the book Zero Day, but while I was going, the whole book was already translated. So I post here a translation of Marvin L. Minsky’s article on the future and human development. For the tip-off on the articles of Minsky, thanks to MagisterLudi



(Scientific American, Oct, 1994 with some minor revisions)
')

Introduction

“Whoever goes to bed early and gets up early, health, wealth and the mind will make a profit” - Ben Franklin.

Everyone wants wisdom and wealth. Unfortunately, health fails us before we have time to reach them. In the future, to prolong life expectancy, development of mental abilities will need to replace the body and brain. But first you need to understand how Darwinian evolution led us to where we are. Then imagine how the replacement of worn parts of the body can solve most of the health problems. Then come up with a strategy to increase the brain to gain wisdom. In the end, we can completely replace the brain - with the help of nanotechnology. As soon as we get rid of the restrictions imposed by biology, we will be able to choose the duration of life - up to immortality - along with other unimaginable possibilities.

In such a future, the attainment of wealth is no longer a problem, the problem is control. Of course, such changes are hard to imagine, and many researchers argue whether such progress is possible - especially in the field of artificial intelligence. But the sciences needed to make these changes are already in the process of becoming, and it is time to think about what this new world will be like.

Health and life expectancy

Such a future cannot be realized by improving biology. We learned about health and its maintenance relatively recently. Now there are hundreds of drugs for various diseases and ailments. But it does not seem that we have reached the maximum life expectancy. Franklin lived 84 years and no one, not counting myths and legends, did not live twice as much. According to estimates by Roy Walford, a professor of pathology at the UCLA School of Medicine, the average life expectancy in ancient Rome was about 22 years; about 50 - in developed countries at the beginning of the 20th century and now is about 75 years. But it seems that each curve ends at 115 years old. Improving the situation in the field of health for all this time did not affect this maximum.



Why is life expectancy limited? The answer is simple: natural selection favors the genes of the offspring. Their number grows exponentially with an increase in the number of generations - thus an advantage in species with an early onset of the reproductive cycle. Evolution usually does not favor genes that extend life longer than the time adults need to take care of their children. And it may even contribute to the offspring, which does not need to compete with living parents. Such competition can contribute to the accumulation of genes that cause death.

For example, after spawning, the Mediterranean octopus (O. Hummellinki) quickly stops eating and dies of hunger. If you remove a certain gland, the octopus continues to eat and lives twice as long. Many other animals are programmed to die after cessation of reproduction. The exceptions are long-lived animals, like us and elephants, whose offspring learn so much from the social transfer of accumulated knowledge.

People are the most long-lived warm-blooded. How could selection pressure lead to our current longevity, twice as much as other primates? This is due to wisdom! Of all the mammals, our babies are least adapted to survival. Perhaps we need not only parents, but also grandparents to take care of and transmit knowledge.

But even with this knowledge it is difficult to avoid death. Some cases are associated with infections. Our immune system has developed versatile ways of dealing with most infections. Unfortunately, the immune system can take different parts of the body as an invader. Such blindness on her part leads to such diseases as diabetes, multiple sclerosis, rheumatoid arthritis and others.

We can get serious injuries. Accidents, dietary disorders, poisoning, fever, radiation and other effects can alter or deform the molecules inside the cells. Some errors are corrected by replacing defective molecules. But if the recovery rate is too slow, the errors accumulate. For example, when the lens of the eye loses elasticity, we lose the ability to focus and need bifocal glasses - one of Franklin's inventions.

The main causes of death are hereditary. These are the genes responsible for cardiovascular diseases and cancer - the two main causes of death, as well as many other disorders, such as cystic fibrosis and sickle cell anemia. New technologies should be able to prevent such violations by replacing genes.

Worst of all, we suffer from deficiencies inherent in our genetic system. The relationship between genes and cells is mediated, there are no drawings on which genes build or restore our body. As soon as we learn more about the genes, we can correct or postpone the onset of age-related changes.

Most likely aging is inevitable for all organisms. Of course, some species of fish, turtles and lobsters show no signs of aging with age, they die from external causes - predator attacks or starvation. Nevertheless, we have no records of animals over 200 years old, but this does not mean that there can be no such animals. Walford and other scientists believe that a specially designed diet, limited in calories, can significantly increase life expectancy, but alas not prevent death.

Biological wear

As soon as we learn more about the genes, we can correct or postpone the onset of age-related changes. But even if there was a cure for every disease, it would still have to solve the problem of "wear." The normal functioning of the cell includes thousands of chemical processes, in each of which errors sometimes occur. The body uses many methods of error correction, but errors occur in many different ways, so no low-level system can correct them all.

The problem is that our genetic systems were not designed for long-term maintenance. The link between genes and cells is extremely indirect; There are no drawings or maps on which genes build or repair the body. To restore defects on a large scale, the body will need some kind of catalog, which indicates which cell types should be located there. In computer programs, it is easy to create such redundancy. Many computers maintain backup copies of their most important "system" programs and regularly check their integrity. However, no animals evolved as schemes, apparently because such algorithms cannot be supported by natural selection. The problem is that the correction of errors will then stop the mutation, which ultimately slows down the evolution of the descendants of the animal so that they can not adapt to environmental changes.

Can we live for several centuries just by changing a certain number of genes? In the end, we now differ from our evolutionary relatives, gorillas and chimpanzees with just a few thousand genes - and yet we live almost twice as long. If we assume that only a small part of these new genes caused an increase in life expectancy, then perhaps no more than a hundred of such genes were involved. However, even if this turned out to be true, it would not guarantee that we could get another century of life by changing another hundred genes. Perhaps we will need to change only a few - or perhaps much more.

Limitations of Human Wisdom

I think even before our bodies wear out, we will face the limitations of our brain. As a species, we seem to have reached a plateau in our intellectual development. There is no sign that we are getting smarter. Was Albert Einstein a Better Scientist than Newton or Archimedes? Is any playwright of recent years Shakespeare or Euripides surpassing? We have learned a lot over two thousand years, but still have not surpassed ancient wisdom, which makes me suspect that we have not made much progress. We still do not know how to deal with conflicts between the goals of individuals and global interests. We are so bad at making important decisions that we are constantly unsure of something.

Why is our wisdom so limited? Is it because we don’t have time to study or we don’t have enough opportunities? Or because, as in the popular myth, we only use part of our brain? Can better education help? Of course, but only until a certain point. Even our best geeks learn no more than twice as fast as everyone else. We have been learning for so long because our brain is terribly slow. This will certainly give more time, but longevity is not enough. The brain, like other finite things, must reach certain limits of what it can learn. We do not know what these restrictions are; perhaps our brain can continue to learn for several centuries. However, in the end, we will need to increase its potential.

The more we learn about the brain, the more we can improve it. Each brain has hundreds of specialized parts. We know only a little about what each of them does, but as soon as we learn how a part works, researchers will try to develop ways to expand its capabilities. They also learn completely new abilities that biology has never provided. As we accumulate this knowledge, we will try to connect them with the brain — perhaps through millions of microscopic electrodes inserted into a large nerve bundle called the corpus callosum — the largest data bus in the brain. With further achievements, any part of the brain will be available for attaching new accessories. In the end, we will find ways to replace every part of the body and the brain — and thus eliminate all the shortcomings that make our lives so short.

Needless to say, in this case we will turn into cars.

Does this mean that the machines will replace us? I do not feel that it makes sense to talk about "us" and about "them." I share the opinion of Hans Moravec of Carnegie Mellon University, who proposes to consider these future smart machines as our own "intelligent children."

Previously, we, as a rule, considered ourselves to be the end product of evolution, but our evolution did not stop. In fact, we are now developing faster, though not in the usual Darwinian way. It is time to start thinking about our new emerging look. Now we can design systems based on new types of “artificial selection”, which can use plans and goals, as well as inheritance of acquired characteristics. It took a century for evolutionists to learn to avoid such ideas — biologists call them “teleological” and “Lamarkian” —but now we may have to change these rules!

Brain replacement

Almost all the knowledge that we accumulate, embodied in various networks within our brain. These networks consist of a huge number of tiny nerve cells and an even larger number of small structures called synapses, which control how the signals pass from one nerve cell to another. To replace the brain, we will need to learn how each of the synapses relates to two cells that it connects. We also need to find out how each of these structures reacts to different electric fields, hormones, neurotransmitters, nutrients, and other chemicals that are involved. Your brain contains trillions of synapses, so this is not an easy task.

Fortunately, we do not need to know everything in detail. If this were the case, our brain would not be able to work. In biological organisms, as a rule, each system has evolved to be insensitive to most details about what happens in the small subsystems on which it depends. Therefore, in order to copy the functional brain, it must be sufficient to reproduce only the necessary function of each part in order to produce its important effects on the other parts.

Suppose we wanted to copy a machine, for example, a brain containing a trillion components. Today we would not be able to do such a thing (even if we had the necessary knowledge) if we had to make each component separately. However, if we had a million construction machines, each of which could build a thousand pieces per second, the task would have taken only a few minutes. In the following decades, new cars will make this possible. Most modern industries are based on the formation of bulk materials. On the contrary, the area of ​​activity called “nanotechnology” is aimed at collecting materials and machines by placing each atom and molecule exactly where we want it.

Using such methods, we could make absolutely identical parts - and, thus, avoid accident, which interferes with traditional machines. Today, for example, when we try to etch very small circuits, the sizes of wires change so much that we cannot predict their electrical properties. However, if we can accurately identify each atom, then these wires will be indistinguishable. This will lead to the emergence of new types of materials that modern technology can not do; we could give them tremendous power or new quantum properties. These products, in turn, will result in computers being as small as synapses, having unprecedented speed and efficiency.

As soon as we can use these methods to create a universal assembly machine that works on an atomic scale, further progress should be faster. If such a machine took one week to make a copy of itself, then we could have a billion copies in less than a year.

These devices transform our world. For example, we could program them to make efficient solar energy collection devices and apply them to nearby surfaces for energy. Thus, we can grow micro-plant fields in much the same way as we now grow trees. In such a future, we will have few problems with achieving wealth, the difficulty will be to learn to manage it. In particular, we must always be careful when dealing with things (like ourselves) that can reproduce themselves.

The limits of human memory

If we want to consider the possibility of increasing the brain, we need to understand how many people know today. Thomas K. Landauer of Bell Communications Research reviewed many experiments in which people were asked to read the text, look at the pictures and listen to words, sentences, short fragments of music and meaningless syllables. Then they were tested in various ways to find out how much they remember. In none of these situations, people could not learn, and then memorize, more than about 2 bits per second for any length of time. If you could maintain this speed for twelve hours every day for 100 years, the total amount would be about three billion bits - less than what we can store today on a regular 5-inch CD. In about ten years, this amount would correspond to a single computer chip.

Although these experiments are not very similar to what we do in real life, we have no convincing evidence that people can learn faster. Despite popular legends about people with a “photographic memory”, no one seems to have mastered, word for word, the content of just a hundred books - or one large encyclopedia. Shakespeare’s complete works are about 130 million bits. The Landauer Limit implies that it will take a person at least four years to memorize them. We do not have reasonable estimates of how much information we need to perform such skills as painting or skiing, but I see no reason why these actions should not be equally limited.

It is believed that the brain contains about one hundred trillion synapses, which leaves enough space for several billion bits of reproducible memories. Someday it will be possible to build a repository for such a pea-sized volume using nanotechnology.

Future intelligence

As soon as we learn what we need to do, our nanotechnologies should allow us to create organs and brain replacements that will not be limited in their “real time” pace. Events in computer chips already occur millions of times faster than in brain cells. Consequently, we could design our "minded children" millions of times faster than we. For such a creature, half a minute will be a year, and every hour will be a whole human life.

But can such machines exist? Many thinkers firmly argue that machines will never have such thoughts as ours, because no matter how we build them, they will always have a lack of a vital ingredient. They call this essence different names, such as feeling, consciousness, spirit, or soul. Philosophers write entire books to prove that because of this lack of a car, they can never feel or understand what people do. However, each proof in each of these books is erroneous, suggesting, in one way or another, what it is trying to prove - the existence of some kind of magical spark that has no detectable properties.

I am intolerant to such arguments. We should not look for any missing part. Human thought has many components, and every machine we have ever built lacks tens or hundreds of them! Compare what computers are doing today with what we call "thinking." It is clear that human thinking is much more flexible, resourceful and adaptable. When an error occurs in a modern computer program, the machine will either stop or produce some incorrect or useless results. When a person thinks, everything constantly goes wrong, but it rarely upsets us. Instead, we are just trying to do something else. We view our problem from a different point of view and move on to a different strategy. The human mind works differently. What gives us this opportunity?

On my desk is a textbook about the brain. Its pointer consists of approximately 6,000 lines that belong to hundreds of specialized structures. If you accidentally damage some of them, you may lose the ability to remember the names of animals. Another injury may leave you unable to make any plans for further action. Another type of violation can lead to the fact that you will suddenly utter curses because of the damage to the area, which usually censor such expressions. Of the thousands of similar facts, we know that the brain contains a variety of mechanisms.

Thus, knowledge is presented in various forms, which are stored in different areas of the brain, and are used by various processes. What are these views? As for the brain, we do not yet know. However, in the field of artificial intelligence, researchers have found several useful ways to represent knowledge, each of which is better suited for some purposes than for others. Most popular use the “If-Then” rule sets. Other systems use structures called “frames” that resemble completed forms. Another program uses networks or schemes that resemble tree scripts. Some systems store knowledge in language sentences or in expressions of mathematical logic. The programmer runs any new program, trying to decide which presentation will best accomplish the task. As a rule, a computer program uses only one representation, and if it fails, the system breaks down. This shortcoming justifies the view that computers do not really understand what they are doing.

But what does it mean to understand? Many philosophers have stated that understanding (or meaning, or consciousness) should be a basic, elementary ability that only a living mind has. I think this is “envy of physics”, that is, they are jealous of how well physics has explained so much with just a few laws. Physicists have acted very cleverly, rejecting all explanations that seem too complicated, and looking for simple ones instead. However, this method does not work when we are dealing with all the complexity of the brain. Here is a summary of what I said about understanding in my book, The Society of Mind. “If you understand something in only one way, then you really do not understand anything at all. This is because if something goes wrong, you are hung up on thoughts that just linger in your mind. Something takes on meaning for us, then, when we associate it with all that we know. Therefore, when someone "cracks", we say that he really does not understand. However, if you have several different approaches when one approach fails, you can try another. Of course, creating too many unintelligible connections will turn the mind into a mess. But well-connected approaches allow you to think, considering things from many points of view, until you find one that works. This is what is meant by thinking! ”

I think this flexibility explains why thinking is simple for us and difficult for computers at the moment. In the Society of Mind, I assume that the brain rarely uses only one approach. Instead, it always runs several scripts in parallel, so that several points of view are always available to it. In addition, each system is controlled by other, higher-level, which track their performance and, if necessary, reformulate the task. Since each part and process in the brain may have flaws, it is hoped to find other parts that are trying to detect and correct such errors.

To think effectively, you need several processes that will help to describe, predict, explain, abstract and plan what your mind should do next. The reason why you can think so effectively is not because we are endowed with mysterious brilliant talents, but because we use agencies of agencies that work together to avoid a dead end. Once we know how these societies work, we can also put them in computers. Then, if one procedure in the program is mistaken, another may suggest an alternative approach. If you saw that the car is doing this, you would probably assume that it is conscious.

Ethics failures

This article talks about our rights to have children, change our genes and die if we want to. No popular ethical system, be it humanistic or religious, has shown itself to be able to withstand the challenges that are already facing us. How many people should be on Earth? What should be the people? How should we share the available space? It is clear that we must change our ideas about how many have children. Conception occurs by chance. However, someday children can be “composed” according to certain desires. In addition, when designing a new brain, it does not need to start from scratch, having so little knowledge about the world as babies. What things should our intelligent children know? How many of them should we have - and who defines their features?

Traditional systems of ethical thought are focused mainly on individuals, as if they were the only value. Obviously, we must also take into account the rights and roles of larger-scale creatures, such as super-humans, what are called cultures, and great, growing systems, called sciences, that help to understand other things. How many such entities do we need? What kinds do we need most? It is necessary to be careful with those who stood in a certain form and resists further growth. Some of the future options are never to notice: imagine a scheme that could consider both your and my mind, and then compile a new unified mind based on our joint experience.

Regardless of what the future may bring, we are already changing the rules that created us. Although most of us will be afraid of change, others will certainly want to avoid current restrictions. When I decided to write this article, I tried these ideas in several groups and answered polls. I was amazed to find that at least three quarters of the audience seemed to feel that our lifespan is already too long.“Why would anyone live for five hundred years? Isn't that boring? What if you outlive all your friends? What to do with all this time? ”- they asked me. It seemed that they secretly feared that they did not deserve to live so long. I find it quite disturbing that so many people have come to terms with death. Are those who feel that they have nothing to lose?

My fellow scientists have not shown such fears. “There are so many things that I want to know, and so many problems that I want to solve, that even ages are not enough,” they said. Of course, immortality would seem unattractive if it meant infinite weakness, weakness and dependence on others, but we assume a state of perfect health. Some people expressed serious concern - that old people must die, because young people need to develop their own ideas. However, if it is true that we are approaching our intellectual limits, then such an answer is bad. We will be cut off from ideas in the oceans of wisdom beyond our comprehension.

Do robots inherit to earth? Yes, but they will be our children. We owe our minds to the lives and deaths of all creatures that have ever been involved in a struggle called Evolution. Our task is to see that all this will not end in vain.

Further reading

Longevity, Senescence, and the Genome, Caleb E. Finch; Univ of Chicago Press, 1994
MAXIMUM LIFE SPAN.1983. Roy L. Walford. WW Norton and Company,
THE SOCIETY OF MIND, Marvin Minsky. Simon and Schuster,
MIND CHILDREN Hans Moravec, Harvard University Press, 1988.
NANOSYSTEMS, K. ​​Eric Drexler. John Wiley & Sons, 1992.
THE TURING OPTION, Marvin Minsky and Harry Harrison. Warner Books, 1992.

CDPV from here , Yokohama kaidashi kikou.