Black technology! The AI has learned to lazy lie?
 
In the field of artificial intelligence from the beginning, always make people feel some worries -- computer will become like a human with consciousness and feelings of the day, and beyond human beings?


Scientists at the Stanford University and Google found that this concern seems to make it more basis.

In 2017, the scientists try to use a CycleGan AI program, the satellite map dubbed said common map.

In order to measure the work of AI, the scientists will further require AI to map again restore the satellite map, were compared before and after.

And it is in this reduction process, scientists were surprised to find that CycleGan can restore a lot of details from the map, let the satellite map and the original reduction almost as like as two peas.

Pictured above, the left is the original satellite map, the middle map, and the right is based on the sketch map reduction satellite map.

On the left the original white building below the "black spots" for example (shown in red circle). These "black spots" of the body is installed on the roof of the facilities, these facilities in the map are not reflected. However, AI can in the reduction, these should have been filtered out to restore the "black spots".

Arguably, dubbed map when filtering out information should not be restored. Therefore, the scientists of this situation greatly puzzled.

They can not intuitively grasp the learning process of AI, so only through repeated testing, trying to find clues. But after the exploration, they finally understand.

Originally, the CycleGan "know" each scientist will make it with sketch map reduction, so in the process of translation, it does not remove those unwanted information, but by some human beings can not distinguish the similar color, the information hidden in the "filter" map.

Although humans can not detect, but the computer can easily distinguish between different colors. So, AI will also be able to "map" from the reduction of large details.

In order to work becomes simple, the CycleGan AI is not in the lazy lie?

Scientists pointed out that the matter should be dialectical view!

Many times, we just added AI to human consciousness and emotion, that it "intentionally lazy". In fact, only AI is given in the task, to find a new way of human not to think. It is not the purpose of "lazy lie", but to "loyal".

But the CycleGan AI should think first will not need information hiding this way, you learn "steganography, which did make scientists feel very interesting.

At present, the development of AI is still in the initial stage, I believe that with the progress of science and technology, they will become more intelligent. But to say that AI has the human subjective consciousness that may not achieve at least a short period of time.

The so-called "AI a", "lazy", in fact, most of the time are just the set of rules is not tight, did not think of all possible conditions, but the AI solution is that we did not expect to place.

These scientists have found their written called "CycleGAN a Master" Steganography of test report.

If you want to know our products, direct contact:

The main business of the company : Lasting machine, Before the machine. , After the machine to help , Heel nailing machine , Pressing machine , Toe forming machine , In the end of the injection molding machine , On glue machine. Italy, after the help of machine, machine before help, help Italy machine, Italy after the help of machine

hotline : 086-0769-88325138 / 13602310875