How to teach AI to distinguish black from white, apple from orange?

Question

Machine learning

Our brain builds associations based on life experiences.
After all, dark and light differ in that nothing is visible in the dark, and not in that darkness has hex number # 000000
How to teach a machine to distinguish and understand what an apple is if it cannot eat it?
Simple example without visual

Он ел не свежее яблоко, поэтому свежий апельсин ему больше понравился.

How to understand what "fresh" is, etc. from the AI ​​point?

Answer 1

After all, dark and light differ in that nothing is visible in the dark, and not in that darkness has hex number # 000000

---- they just differ in that after scanning the image for further processing, including programs that implement what is commonly called AI, the colors take on different numerical designations. And more - from the point of view of computer processing - they just do not differ in anything.

How to teach a machine to distinguish and understand what an apple is

--- so "distinguish" or "understand". Distinguish - see what is called "methods of recognition of visual patterns". Today it is already a purely technical, almost routine task. Dozens of implementations and hundreds of thousands of sources on the net, where everything is described in detail. "Understand" - and no AI actually "understands" anything. If you don't believe me, then give a definition of what you mean by the word "understand".

Simple example without visual

--- and what is not clear here. And why does AI need to understand what "fresh" is?
In general, do not confuse sci-fi articles and movies about artificial intelligence and what "sort of AI" really is today.