AI and Botts chat sometimes are called user computers for words-for artificial intelligence, because in fact, not being aware of anything. He simply conducted calculation and gave results based on statistical analysis. Technically, this is completely true, but why can anyone imitate people very well? Portal scienceealert.com I found it In the problem.
The similarity to the computer is properly criticized, because it hits the most common aspects of AI generation. Unlike the conversations, computers have no integrated prejudices, they are not confused and not a basic moral problem. But the most important thing is that the Viking's calculations are made by those who are specially designed in the way of adhering to the daily language of everyone.
Most people who say this or that language only guess that their interaction with the outside world is a product of statistical calculations. For example, remember the discomfort that you feel when someone says Pepper Pepper and Salt, instead of salt and pepper. Or the strange look if someone nearby ordered strong tea instead of strong tea.
Rules for adjusting order and selecting specific words, as well as many other languages, comes from our social frequency of social interaction with them. The more we often hear something that is pronounced in a certain way, seemingly less acceptable to other alternative options. More precisely, the less likely any other calculated string.
In linguistics, such chains are called “collection”. They are only one of the many phenomena showing how people calculate the magnetic patterns, starting from intuition. We choose the expressions based on them that seem suitable, natural and human.
One of the main achievements of large language models (and conversations) only when normalizing this factor. AI generation – nothing more than the most powerful arrangement systems in the world. Hopefully the statistical relationships between notification codes, whether it is words, symbols or color dots, in an abstract space that marks the meaning and meaning, who offers a text that is not only easy to imitate a person. He can make a person who truly loves artificial intelligence.
Why is it possible? Because AI generally stems from linguistics, often forgotten in the context of global technological reporting. AI tools are a product of not only computer science, but also one of the branches of linguistics. The ancestors of the current language models are tools to translate the machine of the Cold War, planning to translate from Russian into English. But gradually the purpose of such systems has shifted from a simple translation to decoding the principles of natural speech processing (meaning humans).
The process of developing linguistic models gradually, starting with efforts to mechanize language rules through statistics, which measures the frequency of occurrence of certain formulas based on limited data sets. Actual models use nerve networks to create a liquid language, but the logic of probability calculation remains the same. It is not important that the scale and form of these calculations have developed extremely.
So why don't people realize this cold math truth? Partly due to the way IT companies promote their products. Whoever never calculates language – they think, think, look or even dream. Accused of artificial intelligence, decoding human language, getting access to human values. But at least today, this is not like that.