9 naive questions about artificial intelligence

2024 Author: Malcolm Clapton | [email protected]. Last modified: 2023-12-17 03:44

Separating fantasy from a realistic future.

What is AI?

Mathematicians, programmers, futurists and philosophers are struggling to define artificial intelligence (AI). On the one hand, it is a large-scale interdisciplinary area of research at the intersection of mathematics, linguistics, programming and psychology.

On the other hand, AI is a mathematical algorithm that creates intelligent computer programs. They are so named because they can perform functions that are considered to be the prerogative of a person. For example, write a poem or music, have a conversation.

AI is created to solve specific problems - all such systems are highly specialized and cannot perform several functions at once. If a program is designed to transcribe human speech, it can never play a card game.

It is customary to distinguish between weak and strong AI. When companies claim to have created a product with artificial intelligence, they mean its weak options: these are autopilots, voice assistants, translators. Reasoning about a strong AI, the one that can think and be aware of itself (that is, in fact, become on a par with the human mind), remains a scientific and philosophical discussion.

A weak artificial intelligence, when translating a text, replaces some words with others according to a certain algorithm, and a strong one can independently understand the meaning of a sentence. This is the main difference.

Are robots also AI? What about chat bots, machine learning, neural networks?

No, we are talking about very close and interdependent concepts, but still they are not the same thing. Let's get back to the definition: AI is a large-scale interdisciplinary field of research (like geography).

There are several specialized areas of knowledge in this area, one of which is machine learning. Along with it, there are natural language text processing, virtual assistants, and recommendation systems. It's like physical, economic, or social geography.

We go down another step lower. A neural network is one of the subsections of machine learning, a mathematical algorithm with automated parameter tuning. In total, there are four main subsections (methods) of machine learning: classical, reinforcement, ensemble methods, and neural networks. Imagine that this is the geography of the oceans in the section of world physical geography.

And where, in this case, are the robots classified? Chatbots, robots, automatic translators, scanners - all this is the final result and presentation format of AI technology.

Can I meet AI in real life?

Of course! We use it for a wide variety of tasks. For example, T9 predicts the word you are writing - this AI recognizes the combination of letters and suggests one of the options available in its database. A home robot assistant that responds to voice commands is an example of technology. Siri is also artificial intelligence.

AI can learn?

Yes, he already knows how to learn and improve. For example, Google created the AI Mastering the game of Go without human knowledge, which independently mastered the ancient Chinese board game Go, learning from misses and victories.

However, it is worth remembering that self-learning is not a mandatory feature of artificial intelligence. There are systems that simply perform a certain task very well and in which they have not "entered" the learning function. These include robots working in production and sorting.

How about recognizing emotions?

Yes, there are computer systems that can recognize emotions by facial expressions of a person during a conversation. The program evaluates the position of the key points of the face (eyebrows, eyes, nose, jaw and lips) and compares them with the signs of possible emotions that are written in its code.

Moreover, systems are able to portray emotions using emoticons or emojis. This manipulation is based on very simple logic: basic emotions (joy, fun, resentment) are easy to predict and simulate, relying on trigger words ("thank you", "sorry", "offensive" and others).

And warning the next question: no, AI cannot experience emotions. They were formed in humans in the process of evolution and socialization. A similar response system is present in animals - they experience joy, rage, anxiety, and so on. But only a person supplemented this spectrum with the ability of rational thinking.

Can AI get smarter than humans?

On the one hand, this is a rather meaningless question, because there is no universal scale for measuring intelligence. For example, we know that the heart rate of a healthy person is approximately 60 beats per minute. But how is the mind to be measured? In the number of books read, knowledge of the periodic table or the ability to give answers to any questions? Can a cat be considered smarter than a squirrel, and an eagle smarter than a viper? How do you compare the intelligence of a physicist-astronomer and a surgeon?

There is a popular test for measuring the intelligence quotient (IQ) of Hans Eysenck, but it is categorically impossible to consider it a universal criterion. In humans, the brain works in different ways and is "sharpened" for a particular type of activity. Until there is a criterion that can be used as an absolute indicator, there will be no such rating.

On the other hand, when we say "machines will become smarter than humans," we rather mean that they will become smarter. And the mind is much broader than the intellect, it is formed in the process of life and depends on a billion different factors. So far, scientists and science fiction writers suggest the only possible (but not yet implemented) option in which AI will become more intelligent than a person: if the technology is implemented on the basis of DNA molecules, and not by modeling neural networks.

Can AI measure its own actions?

No. In order to evaluate actions, a person, in addition to the thought process, requires moral attitudes, emotions and cultural norms that change over time. Technology is (at least not yet) not available.

Can AI be hacked?

Yes, it is possible. It is a human-controlled program. Any program can be hacked.

Is it possible that the AI gets out of control and decides to get rid of the person?

Despite the fact that we have seen a lot of stories in which robots become aggressive monsters, block the life of cities, seize classified information and commit other crimes, this is possible only in the movies.

AI performs exclusively those tasks that were laid down in it by the programmer. The technology does not provide for independent goal setting. Artificial intelligence may end up in the hands of criminals and cause harm, but again it's about human will.

A different scenario can arise only if the problem of strong AI is solved. This is not possible today. Therefore, all arguments about the probable enslavement of mankind by artificial intelligence have no real basis.