Lin and Fen worked together, and with a clear goal and a clear target, the progress of magic research was quite rapid. Besides, it's not a very difficult ability to show colors, and of course it's impossible to defeat two people.

It's just that what Lin wants is not to show colors like a slide, so that the strange things in the tower consciousness realm can be absorbed unilaterally. Rather, it is necessary to form an interactive action like a 'dialogue', and the changes of one's own side should be able to change according to the response of the other party.

This is different from machines that are designed to solve magic cubes. Regardless of how many times several Rubik's cubes, the situation is that the problem has been fixed, and after the program is answered, the machine will complete the Rubik's cube according to the steps.

To put it simply, the Rubik's Cube machine has two clear steps. The first step is to check the situation of the Rubik's cube and solve the program; In the second step, the robot arm is running and the Rubik's Cube is completed.

But the Rubik's Cube is so easy to answer because the result of each spin is one-to-one. There will not be a situation where the same configuration of the Rubik's Cube rotates the same part but gets different results.

But the strange thing in the tower consciousness realm is not so gentle. Even after a few simple tests, Lin couldn't determine if this thing really had a pattern. The only thing that is certain is that the 'color' will actually attract the other person.

At this point, machine learning comes in. After each color change, it is necessary not only to record the change of the strange thing, but also to re-judge how the color should change next time.

It can be said that except for the pattern that you want to complete in the end, the other parts are made by learning and doing. It's very difficult to say that this work is difficult, but it's not that difficult to say, but it's not that difficult.

This kind of deep learning gameplay, both of them have experience, and it is not difficult. The difficult thing is the use environment.

There are no eight powers in the Tower Consciousness Realm, and all consumption can only be supported by the mage itself who enters. This means that the time spent on solving the problem should not be too long, the consumption should not be too large, and the computing power should be ......

Forget it, the last one is not a problem for someone. In fact, with the ability of the second-dimensional tower to produce and sell itself, Lin is not afraid of the consumption in the consciousness of the tower. But with that strange thing on the sidelines, someone wouldn't dare to consume themselves to the point of losing their ability to protect themselves.

Another concern is, how many times are there opportunities for trial and error?

In the name of testing, though, I've tried that weird thing a few times to react to color. However, it is very simple to use a few color blocks to trigger a reaction to confirm that there is a situation like 'yes' or 'no', and it is not enough to observe the logic of the reaction.

Nowadays, if you really want to bring the solution method into the machine learning program, you can expect that once you start solving the problem, you can't stop it, because it means that the previous work is lost.

The reason for this is that the strange thing is changing all the time. If the program of solving the problem by itself is interrupted, will it stop changing itself and wait? Or is it just a change, and all the previous efforts have been made in vain.

What is uncertain is that once the induction change has begun, is it possible to reopen it if it is interrupted? If so, is there a limit to the number of times?

There are such concerns because the target is not simply a dead thing. Let's just say that the Rubik's Cube also has an upper limit on the rotation of the material structure, but it can't be played without restrictions.

If there is also an 'emotion' in the target, whoever is provoked again and again will eventually explode. When that time comes, will you be able to accept that result? Is it possible for the target to continue to cooperate after the outbreak?

One example that someone can think of is plants like mimosa and Venus flytrap.

Although they have some kind of trigger mechanism that allows them to make a 'closing' action, they are ultimately just plants and not as flexible as animals. If someone deliberately teases them, they may not play two or three times, and they will strike.

Strikes can be due to the fact that the trigger hasn't reverted yet, or because they've been played so often that they have caused permanent damage and can't do a certain action anymore.

Maybe that's just the limit of what can be done structurally, but why can't we understand this state of affairs as the individuality of the plant? And the plants are so individual, how can they ensure that the monster in the tower consciousness realm is a good baby.

Therefore, Lin can only try to complete the control of that strange thing at one time, instead of repeatedly challenging its patience.

It's just that in the whole process, if there were already AI trained and familiar with programming languages in the past, the job should be much easier now. But all along, someone has never thought of making themselves an AI butler similar to Jarvis.

In terms of hardware, the characteristics of the magic tower make it not inferior to the supercomputers in my hometown, and it can even be said that there is no upper limit on performance, only the waste of performance caused by garbage code.

The device based on the Purple Transformation Grade Magic Stone, although the efficiency is a little worse, it is far better than those server-level computers in my hometown.

But even with such an advantage in hardware, someone has never thought of making a set of artificial intelligence. At most, it is a weak artificial intelligence that is born specifically for a certain topic, and carries out limited work.

Like Jarvis in the movie Iron Man, he is just a very powerful weak artificial intelligence at best. While many things can be done, they can only be done within the limits of what has already been demarcated.

Ultron and Vision, the type that can make independent judgments and make their own conclusions, barely constructs the edge of strong artificial intelligence. Skynet in Terminator is even better.

Strong artificial intelligence in the real sense is like the matrix in the hacking mission, which incorporates all the actions of the pros and cons into its own plan, and is a type that cannot be broken by human intelligence.

In other words, it must reach a level that the human brain cannot understand, and it can be regarded as a strong artificial intelligence that truly crosses classes.

It's as if in the research of biologists in my hometown, chimpanzees are the closest creatures to humans, but they only have the intelligence of young children. And the knowledge of higher mathematics cannot be understood by chimpanzees even if they are placed in great detail.

This kind of performance can be regarded as crossing the wisdom gap of classes. And this is also a topic that only exists in theory before a certain crossing is made, and has not yet been realized.

In many science fiction works, the development of strong artificial intelligence is pessimistic. In short, strong AI will not be resigned to the perspective of a servant, but will become a rational ruler.

A little better, lead a group to develop. For example, the Kree Empire in the Marvel series is supreme intelligent.

The worst is the enslavement of a group for a purpose that humans do not understand. For example, the matrix in a hacking mission.

The most ruthless is Skynet in the Terminator, which aims to destroy humanity. Regardless of its purpose, it is necessary to destroy the parasites on the earth that constantly consume resources; Or simply look at human behavior and feel uncomfortable.

But anyone who is playing with AI knows that the performance of AI depends on what is done during the training phase. If you go the wrong way, there is no other way to remedy it than to cut it down and retrain.

The most well-known example is Microsoft's conversational AI. It has only been open to the public for two days, and it has been trained by a large number of malicious netizens to become swearing-mouthed racists. forced Microsoft to shut down urgently and did not dare to open to the public again.

In fact, everyone in the know knows what to do with the abandoned artificial intelligence. There is no other way to save it than to overwrite the old save. And it's not like a game save, the file overwrites the past, and the operation in it is more complicated.

This means that the AI team hired with a high salary has been in vain for a while. If you describe it as a eunuch from the palace, even if you don't cut it all off, you cut off a lot of it.

With someone's ability to educate children, they don't dare to guarantee that they will be able to educate their children into talents, let alone play with artificial intelligence technology. Lin Ke never thought that he was a technical master, he just relied on the knowledge that he didn't have to bully people.

If this knowledge is familiar, then your advantage is gone. It's like Fen, who has mastered a considerable degree of knowledge of the earth, but eats a certain traversal to death.

This is also the reason why after being backstabbed by the children of the academy during the period of the Southwest Peninsula, Lin kept a certain distance from the children of the elementary school.

It's the same for outsiders, and someone is like having a panic disorder about having a baby. It's not that your own children can be abandoned if you abandon them, and you can stay away from them if you keep your distance.

If you have to be responsible for it to the end after you give birth, it is better not to give birth in the first place. In the same way, if you come up with a set of artificial intelligence with endless troubles, it is better not to do it in the first place.

Especially if someone is not confident in themselves. Tony, then? Stark's kind of technical bull has become an out-of-control Ultron. Lin Ke doesn't think that if he does it himself, he can do it better. Don't pit others before you can trap yourself.

So all this time, someone never wanted to make much progress on artificial intelligence. At best, when necessary, get some targeted weak AI to top.

Besides, the place where I am is is a magical world. This world has souls, there are elemental spirits, and there are many things that are difficult for science to explain.

For a magician, there is a need for a trustworthy servant who specializes in handling simple things, and there are many ways to meet this need besides recruiting an apprentice.

For example, stuffing a soul into a puppet. While this is an evil practice, combined with a soul-torturing device, it can ensure that the puppet will not betray and will possess human-level intelligence and know how to do things satisfactorily.

Or summon elemental elves and demons, all of which can meet the same requirements. For these methods, magicians who are already familiar with them naturally have a set of mature methods to avoid most of the pitfalls.

With such a convenient technology in front of him, of course, a certain crosser does not need to deliberately delve into artificial intelligence technology. What he has always done, magic and science are used to complement each other, which is convenient and which one can be used by himself, then choose which.

Obsessed with earth technology can crush everything, and once the nuclear bomb comes out, whoever competes with it must also be able to rub the nuclear bomb in hand. If you can't grow to the realm of hand-rubbing nuclear bombs, and you still have to adhere to the theory of the supremacy of science and technology, then whether you can survive to the later stage is a question.

In short, practicality is more important than any persistence in the face of the subject of survival. This is the way of survival of a small person like him, not a superstitious technology.