Rebirth of the Game Pioneer

Chapter 112 Different Machine Learning
People often make a judgment based on past experience and cognition.

So what is the machine made of?

For ordinary computer algorithms, the algorithm makes a judgment through programming language.

Here is a simple example using if and else statements (for the sake of simplicity and clarity, strict programming language is not used for judgment conditions and output, please don’t mess with it).

if (reader x voted for recommendation):

Reader x is very handsome
else:

Reader x looks average

For ordinary computer algorithms, through the above judgment statement, it knows that the readers who voted for recommendation are handsome.

But if the programmer discovers one day that some readers who voted for recommendation seem to be quite average-looking, then he optimizes the algorithm and changes the judgment conditions to (reader x voted for recommendation and reader x voted for monthly vote).

At this time, for the algorithm, only readers who voted for both recommendation votes and monthly votes are handsome.

Later, the programmer will find that no matter how he adds judgment conditions, there always seem to be one or two readers who will jump out and overturn his conclusion. It is very difficult for him to manually formulate rules to allow the computer to accurately identify handsome readers.

At this time, machine learning algorithms need to be used.

So how do machine learning algorithms work?
We only need to randomly select a certain number of readers (called training data in the field of machine learning) and make a table, which records some attributes of these readers, such as whether they have voted for recommendation votes, whether they have voted for monthly tickets, whether they have left comments, Whether it has been liked, whether it has been rewarded, etc. (called features in the field of machine learning), record the conclusion in the last column, that is, whether it is handsome or not (called tags in the field of machine learning).

Provide such training data to the machine learning algorithm. After training, it will learn a model about the relationship between the characteristics of the reader and whether the reader is handsome.

At this time, let it judge whether a new reader is handsome, and he will give a probability that the reader is handsome and a probability that he is not handsome based on this training model.

Obviously, as long as the larger the amount of training data for machine learning, the higher the probability that it will draw correct conclusions after learning.

Because of this, machine learning is widely used in many work scenarios, such as directly identifying whether the driver's current driving state is normal driving, smoking, making a phone call, holding the steering wheel with one hand, chatting with passengers or doing other things after the road camera takes pictures. thing.

With such powerful learning capabilities combined with equally advanced artificial intelligence technology, it’s no wonder people are worried about the so-called omnic crisis.

However, there is a big difference between the machine learning mentioned in this paper and the machine learning that Gu Feng learned based on his past life cognition.

Based on the world's more advanced artificial intelligence technology and digital life technology, machine learning can directly teach language and behavior.

This is the same as parents teaching their children what a pig is and how to write.

In fact, in essence, this process returns to the most primitive algorithm, that is, the programmer uses the algorithm to let the computer do something.

It's just that in this kind of machine learning in the field of digital life, the digital life can automatically recognize the words it hears and the phenomena it sees, and then generate a new code on its own. This new code is its response to a paragraph or some kind of message. Cognition resulting from phenomenological learning.

Although the first digital life has not yet been born, people are much more confident about the artificial intelligence born under this technology than the artificial intelligence in previous generations.

After all, under the constraints of digital life, this form of artificial intelligence cannot learn quickly by freely traveling through the Internet. It must also be like a human being. It needs to listen, see, and do things in order to gradually learn and improve itself. Cognition and ability.

Gu Feng gave the scientists a big thumbs up without hesitation.

If such technology is not used in the gaming field, it will be a waste of resources.

In his previous life, as early as more than ten years ago when he traveled through time, there were already many games that used the gimmick of APC to attract players, that is, NPCs that claimed to have AI.

In fact, this is understandable. After all, although players don't particularly care, if the NPCs in the game can be more humane and flesh-and-blood, who would want to face a person who will always only say a few repetitive sentences?
According to the information Gu Feng consulted, digital life technology may be mature enough to be commercially available in just the past two years.

Now that the technical bottleneck has been confirmed to be no problem, the super IP plan in Gu Feng's heart can finally start to plan, and strive to implement it as soon as possible.

Opening a blank document, Gu Feng lightly typed the title on the first line - Pokémon World.

Gu Feng's speed of writing the planning book this time is far less than when he was writing the game design book before. At this time, he is like a web writer who has suffered a lot and hates deeply. Sometimes, I often think about it for a long time before I can come up with a short paragraph.

Gu Feng is not too sure whether the animation or the game of Pokémon in the previous life came first, but it doesn't matter. Since you want to maximize the value of the IP, the game must not be made before the animation.

Manga game and game manga are completely different things.

In addition to animations and games, Gu Feng also considered novels and movies in his plan.

Of course, the movie here does not refer to the animated theatrical version, but to movies like "Pokémon: Detective Pikachu". After all, the animated theatrical version is still targeted at the audience who watch animation, and "Pokémon: Detective Pikachu" is aimed at the audience. Dream: Detective Pikachu" is a second-generation movie that can attract more people who don't like watching animation.

However, the planning of the movie is not particularly necessary, and it does not need to be given a high priority in planning.

As for animation, Gu Feng also has a bold idea. He does not plan to use the original animation.

The reason is simple. Although the original work is a classic, it is still a little too young for the age. Sometimes he even wonders, if the Pokémon animation was born 20 years later, would it still be popular?Maybe not.

As far as the theme of Pokémon is concerned, it can definitely be more mature in terms of story arrangement and character creation. After all, the ultimate goal is a game. In theory, although VR Pokémon is expected to target players of all ages, Gender, but in fact Gu Feng knew that the real core player group must still be young players aged between 18 and 20 years old.

Strong competitiveness, high energy, and high enthusiasm for game research are all characteristics of players of this age group. Since Gu Feng wants to make it an epoch-making competitive masterpiece, capturing the hearts of this group of players is naturally the top priority. .

Therefore, it became an imperative choice to abandon the original work and make the animation more mature.

(End of this chapter)