Artificial intelligence is only in its infancy, but its limitless possibilities and dangers are already visible.
Where artificial intelligence is already used. when we type a word in a search engine on the internet and click on one of the many suggested links, then we increase the authority of the link that we clicked on, and reduce the authority of all other links. Based on this data, the search engine rebuilds the output so that the most popular options are at the top. The search engine can randomly place a link on the first page for a while to evaluate its popularity, and if there are no clicks on it, this means that the search engine made a mistake, based on this information, it will reduce the indicators for this link. In this case, we see an example where a computer algorithm learns from its mistakes, automatically offering better and better solutions. The search engine and its configuration algorithms are an example of elementary artificial intelligence.
Artificial intelligence – is it achievable in principle? At what stage of development is the science of AI
Let’s face it, the AI search engines (Yandex, Google) – it is the mind or not? in fact, the level of development of ai today is the same as the level of development of technology before the invention of the wheel. when there were not only cars and planes, but even carts had not yet been invented, but they learned to put round logs under heavy loads in order to transport them faster. So now, AI is at the very beginning of its development. as for the ability of the machine to compete with the human mind – the debate on this topic does not subside. For centuries, people have dreamed of flying like birds, and for centuries most people have been skeptical of the idea that humans will ever be able to fly. It’s obvious – people don’t have wings! For a long time, planes fly much faster than birds, helicopters allow you to hover on the spot, as can very rare birds, and rockets carry research vehicles to distant planets. People did not completely copy the entire structure of birds to create an airplane-although the most important functionality-the use of wings and aerodynamic forces has many similarities. Also, when creating artificial intelligence, we are not necessarily talking about the fact that a person will 100% copy his own brain. A person gradually understands how intelligence works, as well as how neural connections are arranged, and all this helps to gradually model the simplest intellectual things on a computer. First of all, when they talk about AI, they mean algorithms that can learn.
Does artificial intelligence have intuition? Myths about the logic and infallibility of artificial intelligence.
It is commonly believed that robots are mathematically verified soulless automatons with impeccable logic, who never make mistakes, that they have no intuition, they are not capable of humor, emotions, and even more so of love. ten articles will not be enough to reveal all these topics, but i will note the main thing, all this is a myth. The AI thinks in the first place intuitively, but logical conclusions for AI is not yet available. Intuition is an instant reaction to a situation, without thinking. a person intuitively walks, talks and even drives a car-without doing complex logical calculations. Intuitive thinking is born on the basis of experience, it always has a certain percentage of errors. AI thinks exactly intuitively-by experience. ai makes mistakes and learns from them – as we saw in the example with search engines. Similarly, all higher animals learn intuitively from mistakes. the only species on earth that can think logically, that is, operate with abstract concepts, is man. Only a person can bend their fingers to count the number of bananas. Monkeys and other animals do not understand how you can count bananas in your fingers, that is, in numbers. Similar, abstract, mathematical thinking is not yet available for AI. Like the monkey, the AI can learn to distinguish between ” many bananas” and “few bananas”, but so far the AI cannot learn to count them. More precisely, if we teach neural networks to add a value from two numbers, then the training will take place quickly. Similarly, animals understand that if we add two bundles of bananas, then there will be more bananas at the output and will give an approximate estimate. but if we give the input of the neural network not two values, but a set, and let them all be just zeros and ones, and you need to count the number of ones, then this problem becomes almost unsolvable for the algorithms known to date. I promised a simple article, but there are no rules without exceptions, and the following two sentences are very complex, you can skip them and read more:) Mathematically, artificial intelligence, as well as animal intelligence, is able to make an approximation and search for values close to previously known values, that is, calculate the values of new points in multidimensional space based on the values of pre-defined points close to them. but it cannot independently create an algorithm for calculating points that are very far from the existing ones, and this is what is needed to calculate the number of units in the input vector of values.