AI’s advancement turns slower.

   AI’s advancement turns slower.

Growing accuracy requires more complicated code. That causes a situation where AI’s advancement slows. When developers created some pseudo-AI tool in a couple of hours in the 1980s, those programs required about 10-50 lines of code. Those programs asked “what's your name?” and then they output the name that the user gave. Then they might ask, “Is the sun shining”? And then the user could answer “yes” or “no”. Then the program replied with something that gave good things in the user’s mind. Today, AI algorithms require billions of code lines. And that causes a situation where the advancement slows. 

So, when accuracy grows in program advancement slows. 

AI’s advancement turns slower. When its accuracy grows. And that means the AI follows the line of the limits in mathematics. Term limits mean the equation that’s the curve approaches zero endlessly. But that curve never reaches zero. If we translate this mathematical equation into an AI model, we can say that when AI approaches human level, the advancement slows down. Maybe. The AI will never reach a complete human level for programmers, but it will reach a level that is almost human-level intelligence. So the base elements in the AI are easy to make.

Then, researchers should find something more accurate. At the same time, they must find new, complicated ways to train their AI. And that means there is a need for more complicated algorithms. Those algorithms require more power, more time, and more accuracy. This means that the programmers use more and more time. That developers can  create more complicated code for algorithms. And how they should react. When accuracy grows. The sector of their algorithm can work turn smaller in the same time. When the need for accuracy grows, the speed of programming slows. 

And the next thing is that the error detection must be at a level where the system can be trusted. Another thing. What slows AI’s advancement is the calculation power. The system needs the entire data center for every query. That means the system needs so much calculation power that developers have no money to buy the systems that they need. Complicated code requires high-power, very high-accuracy systems. And when the system requires lots of capacity for the smallest duties. 

That causes a situation where the developers don’t have time to use AI as they need. When some details keep the system busy, it has no time to drive new code. Complicated code requires a complicated- and long-term error-detection process. The human coder cannot check even billions of lines of computer code. The entire human lifetime is not enough for that process if the human coder wants to make that thing without automated tools.