Chapter 98 Artificial Intelligence

In order to avoid data leakage, Tony made all the materials into e-books. The e-books in this format are his original creations. Even if others copy them, they will not be able to view them without his key and software.

The military satellites of country M are all gone, and the military navigation service used by his private jet is gone now, so he can only apply for a civil aviation route according to the normal procedure, and finally boarded the plane after wasting a few days.

It was a long journey, but fortunately, the information was thick enough for him to study for a while.

The 50-page general outline, or the design standards, has long been kept in Tony's mind, and the material he is studying now is the design framework of AlphaGo.

Little Annie's face is good enough, and Cheng Wanli gave him all of it, on condition that it should not be spread outside.

Sadly, Annie turned down his request to be her godfather on the grounds that Tony's reputation as a playboy would make her look bad in front of her classmates.

It was not an easy title to take away, and Tony was devastated when he was rejected.

But Annie also opened the mouth for him. One day his reputation improved, and she agreed.

So, he plans to earn some more money and buy some TV stations...

Cheng Wanli and the three of them persuaded him not to waste energy and talent on some useless things, but if these boring things can be used to obtain coins to buy TV stations, and then change his reputation, and then get Anne's change of mind, then It just makes sense.

There is nothing that I, Tony Stark, can't buy, and if there is, it's because I didn't find the right relationship!

It seems he understands the importance of relationships.

He shook his head, when will I, Tony Stark, need to get involved?

Well, except this time in Los Angeles, I'm Tony Stark... (10,000 words omitted)

But why couldn't I come up with such a general outline?

In fact, Tony can understand this thing at a glance, and it belongs to the series that is worthless.

There is actually nothing wrong with Jarvis's design, mainly because the Jarvis he designed is aimed at assisting scientific research intelligence, and the fuzzy neural network design style like AlphaGo is a natural branch of two styles.

If AlphaGo continues to develop like this, it will become more and more like humans, including its way of thinking, but this does not mean that AlphaGo has no shortcomings.

AlphaGo has shortcomings. First, its precise calculation ability is not enough, and it is not even comparable to Tony's oral calculation ability, let alone its great mathematical ability.

Second, it can forget things just like humans! Note that it is possible, not yes, it is still different from human beings.

In the general outline, the knowledge representation of artificial intelligence is divided into four categories: judgmental, logical, neural network, functional or procedural.

These four design directions of intelligence can be combined into many combinations, and artificial intelligence with different performances and performances can be designed.

Only one kind of development will have many limitations, such as Jarvis, which focuses on judgment and logic, which can help him provide rigorous calculation results. The disadvantage is that the artificial intelligence designed in these two directions, if not good Algorithm, there will be a lack of flexibility, slow response characteristics.

Of course, this direction also has advantages, such as the judgment direction, which is natural, modular, uncomplicated and easy to use, and the generated intelligence has excellent memory and can’t make mistakes, which is a necessary product for scientific research.

The logic direction makes up for the shortcomings of the judgment direction's lack of reasoning, and it can be regarded as a universal assistant.

And the design direction of neural network will produce the results like AlphaGo, it will make people look or touch it and think it is human, or there are humans in the background manipulating it to answer, it imitates the form of human memory, It has the characteristics of fuzzy, associative, fast, and anthropomorphic. Of course, there are many shortcomings, such as imprecise calculations, reasoning ability depends entirely on blind guessing, and so on.

Functional procedural formula is an essential oil, and it can be used in both directions. The advantage is that it can be modularized and can improve the speed. The disadvantage is that it is difficult to design universally.

Knowing this, he has more confidence in Jarvis's design. First of all, he has to choose the direction.

The original intention of Jarvis's design was to assist him in his work. What he needed was an intelligence that could help him deduce, memorize, and precisely operate tools or equipment, so his previous design direction was actually correct. As for why he couldn't recognize it? How many screwdrivers, under the guidance of the general outline, he already knew the optimal solution for this problem.

Including the good mathematical solution that Annie told him before, is actually solving the problem in a judgmental way.

The design performance is: calculate the shape and size of the opening of the screwdriver taken according to the distance from the observation camera, and then calculate the distance and size of the target screw, replace the screwdriver if it does not match, recalculate, and start unloading the screw when it matches.

But if this is designed with a neural network, the design is as follows: I have three flat-head screwdrivers on my body, which are No. 3, No. 5, No. 7, and put them in the three card slots in order. There are cross and plum blossoms beside them. . The target screw looks like a small slotted screw, and it looks like a size 3. I will try the smallest slotted screw first. Removed perfectly! I'll just say I'm a mechanical genius.

The judgment method can adapt to all tools, regardless of the number, you can always try out a suitable one, but it takes a long time. The network method can only adapt to known standard tools. When encountering non-standard parts, unless you let him learn new standards and add new tools , otherwise it will be there stupidly, just like a fool.

These two directions are not to say that there cannot be intersections. On the contrary, they can be intersected in an appropriate amount to create an artificial intelligence that is more suitable for you.

Therefore, after he studies the design structure of AlphaGo and refers to its design concept, Jarvis's intelligence can make a qualitative leap.

He originally wanted to learn a little Chinese, but for some reason, Pepper told him to give up, saying that he would find a better way to learn it later.

He has already seen the simplicity and convenience of Chinese in describing technical articles. Moreover, Cheng Wanli is right. Learning a few more languages ​​and understanding the way humans think and behave in various languages ​​will help improve the level of artificial intelligence.

He agrees with that.

The scenery on the ground outside the window also gradually turned into terracotta and yellow sand. They were getting closer to their destination, and he didn't even notice when they stopped to refuel in the middle!

The plane landed soon. This was not in their capital. Because they carried experimental weapons, they landed at the airport in the US-controlled area, and then went to the designated location.

The receptionist given by the military is his good friend Colonel Rhodes.

"Hi, Tony!"

Thank you for being too lazy to name, how can you repeat the reward!