AI can make many things better than humans.

AI can make many things better than humans. The operational sector of AI is narrower than humans. But if the thing that the system should make requires precise multiple calculations, the AI is better than humans. The thing is that AI can make things like bridges, ships, and aircraft faster than humans. In those cases, the AI handles the precise information. And the system can calculate things like flows on the aircraft's surface very carefully. 

That kind of simulation requires a large number of precise and hard calculations. So when the AI-based system develops the virtual model in its memories it can discuss or exchange information with the simulator that compiles with parameters like flow cross sections, friction, radar cross sections, and other things. 

The problem with AI is that it doesn't know what it does. The AI can play chess or some card plays better than humans. When the AI plays chess it simply simulates the game and calculates possible movements of the buttons. 

When the AI plays cards by using a robot hand and physical cards, it simply uses image recognition. Every card has and card combination that has a certain number value. The bigger value wins the smaller value. The system calculates the possibilities of the card combinations. And of course, it must know that it would not change the wrong cards in poker games. The system can observe faces, changes in voice, and movements of the people for estimating the possibility that the opponent player has a good hand. Or the AI can search reflections from the glasses of the player. 

AI can make many things better than humans. And in the future humans must explain where they are better than AI. When we think of the AI as an artist and look at the image above this text, we can see that the human characters are somehow bizarre. We can see that the image is somehow different than natural things. So we could say that it's abstract art. 

But if we want to make the AI make things like a novel. It makes it by connecting data from different sources. And there is the possibility that the result is also an abstraction, a nonsense series of paragraphs full of texts. But that text has no connections with topics. That is the reason why people who want to give AI to make their thesis should read the product before they are delivering it forward. 

The reason is that the AI does not know what reads in the texts that it uses. The AI would use the search engine and then select sources from the search results. But the problem is that the AI would simply select the first paragraph from the first result, the second paragraph from the second result, etc. And that thing makes the result something, that will seem fine at first look. But then the text has no sense at all. 

The reason why this thing happens is that the AI doesn't understand the text itself. It uses search results for making things that might look like real things. Of course more advanced AI can have parameters that should find in the paragraphs that they use for developing new text. But in that case, the AI uses the elements that it connects to the entirety.