Dmitry Volkov is the founder of the SDVentures holding a PhD in Philosophy, author of a series of works on the philosophy of consciousness and co-founder of the Moscow Center for Consciousness Studies at MSU. He told the participants of the LAKHTA VIEW session about the issues of artificial intelligence and consciousness and shared his vision on how can we coexist with superintelligence.

Machines that today have rudimentary expertise in different fields, will eventually learn to think about general things.  After a while, these machines will be more intelligent than people. Philosophers (Nick Bostrom, Raymond Kurzweil and others), and even developers, for example, Elon Musk, talk about this more and more. These experts say that superintelligence is possible, and artificial intelligence is the way to it. I will introduce just a few of the possible ways to reach it.

Emulation of the brain

One way to superintelligence is brain emulation. We know that the brain is a colossally complex organ consisting of approximately 100 billion neurons. All of these neurons are connected. This is obviously a complex structure, very difficult to emulate, but not impossible. One way to do this is by using machines like the most powerful supercomputer in the world, the Sunway TaihuLight.  The other way is to create a simple machine that can be trained. The machines are already self-learning. If we continue this way, we will have machines living with us and developing their intellect. This trend will not develop consistently, but will accelerate all the time.  Thus we can reach the intellectual point of singularity (when in a finite amount of time the development of intelligence will be infinite). If it is possible, what would it mean for us?

Dmitry Volkov
Dmitry Volkov

Philosophers wonder whether our ultimate goals focus on intelligence. The fact is we cannot measure machine intelligence with the same criteria we use to measure that of a human. We have to understand that artificial intelligence will not evolve as human intelligence had. It can be, that there is no connection between intellectuality and the ultimate goals of AI, that is, perhaps a machine capable of thinking will have completely absurd goals (for example, calculating the number of Pi to infinity). In other words, the ultimate goal and intellect can be unrelated, and this thought helps us to remain calm. But there is another worrying hypothesis, the one of instrumental convergence. It states the following: intelligent agents can allocate a group of instrumental (intermediate) goals (and values) that will be very similar for all of them, and the achievement of which will increase the agent's chances of achieving any ultimate goals from a wide range. It is argued that a variety of intelligent agents will seek to achieve these intermediate goals. Some philosophers believe that intelligence may have different ultimate goals, but the instrumental intermediate goals will necessarily be identical. One of the comprehensive goals is increasing and preserving one's intellect. To achieve any goal, intelligence will need to sustain its intellectuality, so any artificial intelligence (even if its purpose is absurd) is primarily interested in maintaining intellectuality. Among other things, artificial intelligence will need technological development, and it will constantly improve from a technological point of view. For example, if you are talking about a robot, it will try to get more freedom, more opportunities to move around and so on. The third (most disturbing) goal that artificial intelligence will have is an instrumental goal, which is to control resources. This goal could become disastrous for humanity. Even if intelligence has the ultimate goal of calculating Pi, then for its realisation it will need to control all the resources (therefore some believe that ultimately AI will try to control people). This will make people and machines compete for the same resources. This thought makes us doubt whether the creation of a thinking machine is a good idea.

We can reach a very interesting conclusion from the idea of thinking machines: humankind is developing as it had done before - building up technological power, learning more about itself, creating smarter machines, maybe even learning how to computer simulate various phenomena. Since machine production is becoming cheaper, most likely there will be more and more artificial agents. In a hundred thousand years AIs will be smarter than human intelligence. If it is possible, then in the future the ratio of biological intelligence to artificial will be one to infinity. An interesting idea: if it is possible for the future, then it is also relevant for the past. Thus every one of you is, in fact, an artificial agent simply from mathematics, from the ratio of biological to artificial. The paradoxical conclusion that follows from the idea of thinking machines is that in fact we are these machines.

Tatyana Chernigovskaya is a Professor of St. Petersburg State University, Head of the Department of the Problems of Convergence in Natural Sciences and Humanities, Head of the Laboratory of Cognitive Studies with a PhD in Biology, the most famous Russian scientist involved in the popularization of science. In her lecture on LAKHTA VIEW Tatyana Chernigovskaya presented the unanswered questions of humanity.

​Tatyana Chernigovskaya
​Tatyana Chernigovskaya

There are many predictions. All of them not very comforting, but also of little value, because the world is changing at such a speed that it is. I believe that we can have no future without understanding our inner selves. We must have something to rely on, something that will not allow us to collapse internally. It is pointless to reach scientific breakthroughs and create miraculous mechanisms, if later on we all lose our wits. Today we are discussing whether to declare a pandemic of mental illnesses, and this is not a joke. The number of people with a mental disorder is increasing. Thus, the statistics for the US suggests that every sixth citizen is mentally unstable. 

What can stop us?

Prohibition of scientific research will not work: there will always be a rich sponsor for those who can create the incredible. We can only stop ourselves, but that is romanticism. As a biologist, I ask: does evolution have a vector, a goal? First there was a simple nervous system, then complicated, then super complicated, and eventually we had our brain. We accelerated every million years, every thousand, every hundred, every decade, and now we accelerate every minute. We have reached a very high speed and I see the main danger in that. We cannot adapt to the speed of the flywheel that we started. Is our body, our psyche able to withstand this?

What happens in our head?

Each neuron has a huge number of connections with other parts of the brain; it makes no sense to look at all these numbers. Neurons, the main protagonists, are located deep in the glial cells. By the way, the number of neurons increased ten times and the glial tissue itself is also capable of much, for example, it has its own memory. It is not a stable network, but a live one, it is changing. If we say that the neural network is a text, and I say that it is a text, then it is a text that is being constantly rewritten. One neuron itself can do a lot, for example, it can identify a person. However, what are all the neurons doing? This issue is studied by cognitive science, a convergent discipline uniting several sciences at once, among them psychology, neuroscience, linguistics, anthropology, artificial intelligence and philosophy, which is losing ground. Today, philosophers are people with trained thinking; we need them to point out the incorrectness of a question. Artificial intelligence is one of the participants in the system, because it is a model of what we have in real life. Why do we need a model? In order to see whether we have a correct image of what we study.  If the model behaves the way we intended, then we are on the right track.

Why do logical conclusions and mathematical proofs apply to real nature?

Immanuel Kant said that "reason does not draw its laws from nature, but prescribes them to it." Another question that torments me is why can’t we apply mathematics to real nature? Why should I believe it?  Does nature have the mathematics of humans? Does it obey our laws? The brain is in the physical world, it is controlled by physics and mathematics. But what kind of physics - Newton, Einstein or Bohr? A decent answer is all of it. And what kind of mathematics do we apply - Lobachevsky or Aristotle? I ask these questions to everyone, I did not receive answers until recently, when I was told that there is no mathematics that explains the brain. It is not a matter of lack of data or time, such mathematics simply does not exist. Meanwhile, we have a pile of data the size of the Himalayas and do not know what to do with it. The trick is to understand and apply it.

Can we "see" our ethical and cognitive capabilities in the brain?

Everyone hopes that we will find everything in the brain. But will we “see” what we are looking for (ethics, cognitive possibilities)? All modern neuroscience is built upon the search for correlation, that is, we are trying to understand what is happening in the brain when a person sings. All of the articles aim specifically at finding the address, and not answering the question "what for?”. Why search for ethics in the brain? A Soviet surgeon Archbishop Luka said: "I have done many operations on the brain and have never seen a mind there, nor have I seen conscience nor honour."

The idea of ​​one of the greatest modern philosophers Daniel Dennett is that if we examine every cell in the brain, we will understand the whole of it. When I argue with my neuroscience colleagues, they say that now we know little, but there will be devices giving information about each and every neuron in the brain. Why do I need to know about every neuron out of 100 billion neurons, what will I do with this data? I do not need this information. It has to be studied, it is needed for other tasks, but the whole is still larger than the sum of parts. Yuri Lotman once said that one can cut a calf for steaks but cannot assemble it back and science is “making the steaks”. It sorts out all the smaller particles, but what for? Which question are we trying to answer?

Do not believe when you are told that there is a gene responsible for singing, reading and other abilities, which is in that part of the brain or another, it is not true. A huge neural does everything. The doctors will say if we violate the parts of the brain responsible for speech, then the person cannot speak, it is a fact. However, if we forget about medicine, put our head in the MRI and give the brain a task, we will see that every part of it works.

The LAKHTA VIEW program includes eight thematic sessions that will take place in the atrium of the House of Books (Singer House) every third Friday of the month until June 2018. The programme will be devoted to organisation and management of the urban space. The LAKHTA VIEW format combines expert lectures with open discussions. Two experts on the topic of the event moderate dialogue with the audience and viewers of the live broadcast.

Translated by Pavel Vorobyev