Why cars will also have to be hard

Every week, Look At Me publishes an excerpt from the new non-fictor-book, which goes out in Russian. This time we present the book of Nick Bostom “Artificial Intelligence”, which is published by the Publishing House “Mann, Ivanov and Ferber”.

Life in the digital world

In the transition period, the lifestyle of a person living under the Maltese conditions will not necessarily be like one of the models we know (say, on the lifestyle of a hunter, collector, farmer or office worker). Most likely, most people will eke out a miserable existence like a loaner-wrant, who is barely enough for their lives. People will live very poorly, in fact, for only interest or state benefits. But at the same time, they will live in the world of innovative technologies - in the world not only supernight machines, but also drugs against aging and drugs that give pleasure; In the world of virtual reality and various self -improvement techniques. And it is unlikely that all this will be available to the majority. Most likely, real medicine will be replaced by drugs to stop the growth and slowing of metabolism in order to save, since for the mass of people, active life will be impossible (if you take into account the constant reduction of their already minimal income). As the population grows and incomes, people can regress to a state that minimally satisfying the requirements for paying a pension-perhaps it will be a brain with a barely squeamish consciousness, immersed in a container and connected to the supply of oxygen and nutrient liquids that serves machines and which is able to accumulate a little money for reproduction by cloning itself by a special robot-technician.

Even greater frugality will be ensured by modeling the brain, since a physically optimized computing substrate created by the ultrasound can be more effective than the biological brain. However, migration into digital reality will slow down that simulation models will not be able to be considered people or citizens and, accordingly, will not receive the right to a pension or a saving account not taxable. In this case, the niche for people will remain, along with an increasing population of simulation models and artificial intelligence systems.

Most people will earn the miserable existence of a loafer-guarantee

So far, all attention has been focused on the fate of our descendants, whose life can support savings, benefits or wages received at the expense of those who hire people-love workers. Now we switch attention to the entities that we still attributed to capital: to cars that have always belonged to people - cars designed to perform certain functions and able to replace a person in a very wide range of tasks. What will be the position of these work horses of the new economy?

There would be nothing to discuss if all these machines would remain machine guns, simple devices like a steam engine or an hour-long mechanism-there will be a lot of such good in a post-transition economy, but it seems that it is unlikely that anyone will be interested in this soulless set of components. However, if the cars have consciousness - if they are designed so that they can be aware of their exclusivity (or for other reasons moral status will be attributed), then it is important to include them in the world system. The well-being of machines will be the most important aspect of the post-transition period, since they will dominate quantitatively.

Voluntary slavery, random death

The first question suggested: will be owned by capital workers as capital (slaves) or will they be hired for wages? However, upon closer examination, doubts arise that something will depend on the answer. There are two reasons for that. Firstly, if a free worker in Maltese conditions receives a salary at the subsistence level, he does not have funds left after paying for food and other basic needs. If the employee is a slave, then the owner pays for its maintenance, and still the slave does not have free funds. In both cases, the employee receives only the most necessary and nothing more. Secondly, suppose that a free worker was able to somehow secure the income above the subsistence minimum (possibly due to the favorable regulation system). How will he spend this increase? For investors, it would be the most profitable to create virtual “slaves” workers who are ready to work for a salary at the subsistence level. This could be done by copying those employees who have already agreed to such conditions. By appropriate selection (and, possibly, some code change), investors could create workers who will not only prefer to work voluntarily, but also decide to sacrifice all additional income to their employers, if such suddenly appear. However, after transferring money to the employee, they will return to its owner or employer in a circle, even if the employee is a free agent endowed with all legal rights.

Perhaps someone, objecting, will notice how difficult it is to create a car who agrees to voluntarily perform any work or victims of his salary to its own owner. But imitation models should have especially close motives for people. Please note that if the initial problem of control that we considered in the previous chapters seemed difficult, now we are talking about the conditions of the transition period - when, apparently, the methods of choosing motivation will be brought to perfection. Moreover, when it comes to simulation models, then it would be possible to achieve a lot, simply selecting the necessary human characters. Probably, the problem of control will, in principle, will be simplified, if we assume that the new machine intelligence will be included in the stable socioeconomic matrix, already inhabited by other law -abiding supernatural agents.

Therefore, I propose to dwell on the plight of workers' cars, regardless of whether they are slaves or free agents. First, let's talk about emulators, since they are easiest to imagine.

In order for a new man with his professional experience and necessary skills to appear in the world, it will take from fifteen to thirty years. Throughout this time, a person needs to be fed, educated, taught, he will need shelter - all this is great expenses.

The life of many machines will be limited to one subjective day

On the contrary, creating a new copy of the digital employee is as easy as loading the next program into RAM. That is, life becomes cheaper. The company can constantly adjust its labor force under changing requirements by creating new copies - and destroying old ones in order to free computer resources. This can lead to extremely high mortality among machines. The life of many of them will be limited by one subjective day.

There may be other reasons in addition to fluctuations in demand, why employers or emulators will often want to kill (disable) their employees. If for normal functioning the brain emulator, like the biological brain, requires periods of rest and sleep, it can be cheaper to wash the exhausted imitation at the end of the day and replace it with a recorded condition of a fresh and rested imitation. Since such a procedure would lead to retrograde amnesia everything learned during the day, emulators engaged in the performance of tasks that require the formation of long -term cognitive chains will be able to avoid frequent erasures. It is difficult to write a book if every morning, sitting down at the table, you do not remember anything created the day before. But agents who do not so intellectual types of work can be reinforced, and this is quite often done - from a once trained seller or employee serving customers, you will need to “keep” the necessary information for no more than twenty minutes.

Since restarts do not allow you to form memory and skills, some emulators can be placed in a special training environment in which they will be continuously, including during periods of rest and sleep, even if their work does not require long cognitive chains. Under such optimal conditions, some customer service agents could work for many years - and with the support of coaches and performance experts. The best students could be used as “tribal stallions”, that is, millions of fresh copies would stamp every day according to their template. It makes sense to invest in large means in these templates, since even a slight increase in their productivity would provide a noticeable economic effect, being replicated millions of times.

In parallel with the task of teaching employees-scores to perform certain functions, huge efforts will be made in order to improve the technology of their emulation. Success in this direction would be even more valuable than the successes in training individual employees, since improving emulation technology applies to all emulators (and, potentially, to other imitation models too), and not just employed in one specific area. You can direct huge resources to search for computational short paths that allow emulators to create more efficiently, as well as develop neuromorphic and completely synthetic AI architectures. These studies would probably be carried out by emulators launched on a very fast hardware basis. Depending on the value of computing power, millions, billions or trillions of imitation models of the brain of the most insightful human researchers (or their rejected versions) could work around the clock, pushing the boundaries of machine intelligence; Some of them could act magnifunes faster than the biological brain. This is a significant reason to believe that the era of humanoid emulators will be short - very short in star time - and that she will replace the immeasurably superior to their artificial intelligence.

We have already listed several reasons why emulators employers can periodically roll their herds: fluctuations in demand for workers of various activities; Savings at rest and sleep; The emergence of new advanced templates. Another reason may be security considerations. So that the workers' imitation does not bear hostile plans and not weave conspiracies, emulators busy in especially important positions could be launched for a limited time with a frequent discharge to the initial state of readiness.

These initial states to which emulators will be returned should be very carefully prepared and double -checked. A typical emulator with a short life cycle, which was optimized from the point of view of his loyalty and performance, could feel just a good rest the next morning. He would remember that after many (subjective) years of intensive training and selection, he became the best among his classmates, that he had just gained strength on vacation, slept, listened to an enthusiastic incentive speech and peppy music, and now he could not be done to make a maximum possible for his employer.

He is little worried about the thoughts of inevitable death at the end of the working day. Emulators suffering from fear of death and other neurosis are less productive and therefore cannot be selected as a template.

Who will work in the world of the future