Disclaimer
* the text below is written by the author in the spirit of "philosophy of artificial intelligence"
* comments from professional programmers are welcome
Eidos are images that underlie human thinking and language. They are a flexible structure (enriching our knowledge of the world). Eidoses are fluid (poetry), can be reborn (change in worldview) and change their composition (learning is a qualitative growth of knowledge and skills). They are complex (try, for example, to understand the eidos of quantum physics).
But the basic eidos are simple (our knowledge of the world is at the level of a three-seven-year-old child). In its structure, it is somewhat reminiscent of a programming language interpreter.
A regular programming language is rigidly structured. Command = word. Any deviation at the comma level = error.
Historically, this has been driven by the need to interact with machinery.
But we are people!
We are able to create an eidos interpreter capable of understanding not commands, but images (meaning). Such an interpreter will be able to translate into all languages ββof the world, including computer languages.
And clearly understand the statement.
An unambiguous understanding is a trap! He is not! As there is no objective reality. There are phenomena (as philosophical phenomenology says) that our thinking interprets.
Each eidos is an interpretation of understanding, and a purely personal one at that. Two people will complete the same task in different ways! We all know how to walk (we all have the same way of movement), but everyone has a unique gait, it can even be identified like a fingerprint. Therefore, mastering gait as a skill is already a unique personal interpretation.
How then is interaction between people possible? β Based on constant refinement of the interpretation!
Human aerobatics is an interpretation at the cultural level, when entire layers (contexts) of meaning are available by default.
The machine is devoid of culture and therefore context. Therefore, she needs clear, unambiguous commands.
In other words, the system "human-computer-artificial intelligence" is in a closed loop or dead end. We are forced to communicate with machines in their language. We want to improve them. They cannot develop on their own, and we are forced to come up with more and more sophisticated code for their development. Which, as a result, is becoming more and more difficult to understand ourselves ... But even this advanced code is initially limited to ... a machine interpreter (that is, code based on machine instructions). The circle is closed!
However, this compulsion is only apparent.
After all, we are people and our own (based on eidos) language is initially much more productive than a computer one. True, we almost donβt believe in it anymore, we believe that the car is smarter ...
But why not create a software interpreter that captures the meaning of human speech not on the basis of commands, but on the basis of images? And then I would have already translated them into machine instructions (if we really need to interact with machines, and machines cannot do without them).
Naturally, such an interpreter will not grasp the meaning well, at first he will make a lot of mistakes and ... ask questions! Ask questions and improve your understanding. And yes, it will be an endless process of increasing the quality of understanding. And yes, there will be neither unambiguity, nor clarity, nor machine calmness in it.
But excuse me, isn't this the essence of human intelligence?..
Source: habr.com