Daming 1805

Chapter 616 Machine Intelligence Learning Creation System

At the end of the 34th year of the Grand Duke's reign, Zhu Zhudi came to Zhu Jingyuan's residence again and presented him with the latest computer and software system.

"Machine Intelligence Learning to Create Systems".

Artificial intelligence similar to Zhu Jingyuan's previous life is just a relatively complete software system.

Thanks to Zhu Jingyuan's prompts and guidance from the very beginning, Judy Plutonium has been working tirelessly to develop more realistic machine intelligence in the process of developing simulated worlds and various games.

However, in the early days, it was basically only used in games. In the past ten years, other real-life applications have gradually emerged.

For example, face recognition, item sorting, driving assistance, intelligent voice, etc.

In recent years, these important basic capabilities have been continuously upgraded.

At the same time, there have been new breakthroughs in the scope of functions.

It is already possible to chat intelligently with users and control devices using voice.

Articles can be produced based on titles and frames, and images based on key points and elements.

Generate an intelligent speech library based on speech paragraphs, and directly generate speech based on game lines.

You can also repair and transform existing pictures, videos, game graphics and sounds.

More and more colorful intelligent applications are being provided.

At the end of Dagong 34, with the mass production of a new generation of small computers completed, Zhu Zhudi made a summary of these functions.

Judy Pu calls these capabilities collectively “machine intelligence learning creation systems.”

Zhu Jingyuan was quite satisfied after looking at it, but after playing with it for a while, he sighed softly.

Ming's technological level has begun to surpass that of his previous life in more and more aspects, but he has also become really old.

I am 64 years old this year.

Although compared to his peers in the two worlds, his current physical condition is actually very good.

But compared with when I was young, the gap becomes more and more obvious.

There are more and more gray hairs on my head, my body is becoming less and less flexible, and my brain is becoming slower and slower.

Judy Plutonium didn't know why her father sighed.

Just when he was about to ask, Zhu Jingyuan had already given instructions to his servant:

"Call Di Lan to come over..."

Then Zhu Jingyuan began to ask Judy Pu about the training logic of his intelligent systems.

After Zhudi Lan came over, Zhu Jingyuan asked Zhudi Lan to briefly explain to him the status of these procedures.

Then Zhu Jingyuan asked a question directly:
"Di Lan, you are the president of Da Zongbo and the Criminal Court, in charge of culture and law.

"How should we deal with the ownership of materials used in machine learning?"

After hearing this, Judy Lan said casually:
"Erchen thinks it's very simple. Just pay for the authorization directly."

Zhu Jingyuan immediately asked:
"What is the logic and reason for charging for machine learning? Legislation must have convincing reasons.

"If machine learning needs to be charged, then do ordinary people also need to be charged for learning calligraphy and painting works that have been made public?"

Judy Lan said of course:

“You have to pay to study.

“Not to mention the learners’ own consumption in the learning process, such as pen, ink, paper, inkstones, computers and other equipment.

“It also doesn’t talk about the cost of schooling and hiring teachers.

"Purchasing study materials, raw materials for study and study content on your own will cost money."

Zhu Jingyuan pointed at the calligraphy and painting generation software on the screen:

“For the calligraphy and painting works that the creator has published for free in real life or on the Internet, there is obviously no need to pay when others learn from them.

“Does this program, which uses other people’s calligraphy and painting as learning materials, also need to pay for all the calligraphy and painting works used?
“I remember that many calligraphers and painters have published their works for free on the Internet.

"As long as it is not used for profit-making, others are usually allowed to copy and learn from it, and some people will even take the initiative to give guidance."

Judy Lan slightly tidied up the language:

“Father, according to the existing laws of the Ming Dynasty, works that creators take the initiative to make public for free cannot be used by others for any profit-making activities.

“This profit includes direct and indirect profits, as well as actions that are not profitable now but will be profitable in the future.

"Those enthusiastic masters usually don't mind if others imitate their own works first and then create different works in a similar style.

“As long as the other party doesn’t copy their designs, it doesn’t matter if they are used for profit-making activities.

“I think it’s a natural thing to do.

“From a legal perspective, the fundamental reason why there is no conflict is because there is no conflict of interest.

“Creators who make their works publicly available for free actually have a mentality and understanding that is taken for granted and was basically correct in the past.

“First of all, unless others directly plagiarize, they cannot simply profit from their own works.

“Even if others learn their own style and recreate it, they still need to spend a lot of energy to complete the creation.

“The other party has to do it himself, writing word by word and drawing line by line.

“The other person’s creation speed will not be significantly different from my own.

“In most cases, the learners’ popularity and income level will not be higher than those of the original creators.

“Learners also usually respect the original creator and view the other person as a teacher.

“This has the potential to bring a reputation boost to the originator, and possibly even customers who can directly contribute money.

“There is certainly no conflict between the learner and the originator in this case.

“Once the learner and the original author use a highly identical style, the learner’s work is more popular than the original author.

“Then the relationship could quickly decline.

“The style of the learner is similar to that of the original author, and the degree of completion is higher than that of the original author himself, and his popularity will exceed that of the original author.

“Even customers who can provide revenue will become customers of learners.

“They are no longer their own potential customers.

“At this time, the learner will become a promoter, and the original person will become an idea provider or style creator.

“Even on this basis, as long as the learner still respects the original author, the relationship between the two parties will usually not become too bad.

“Because those who carry it forward will provide additional visibility to the original creator who has fallen behind.

“At the same time, the person who carries forward the glory is also a human being, and the speed of creation is limited by the limits of the human body.

“In the event of an increase in orders, original creators with similar styles have the opportunity to obtain additional practical benefits.

"The conflict of interest between the two parties is still not obvious, and the interests are still convergent.

“If a learner invents a machine, he can put the original author’s manuscript into it and generate countless similar high-quality works in an instant.

"And make a lot of money with the 'cheap industrial products' produced by these big machines.

“The original creator will definitely be furious.

"So we don't need any complicated persuasive reasons.

“As long as it is stipulated that machine learning needs to be charged, most of the existing original creators will directly support this regulation.

“Because the machine is likely to prevent them from benefiting from it, they will be quickly replaced by the opponent’s machine and lose most of their living space.

“Like traditional hand weavers versus mechanized textile machines.

“However, the relationship between hand-woven textile workers and mechanized textile machines is not the same as the relationship between calligraphy and painting creators and machine learning intelligent drawing.

“The inventor of hand-woven technology cannot be determined, and hand-woven workers are just workers who master the technology.

“There is no creativity in the worker’s weaving process itself.

“Others invented and purchased mechanized textile machines without direct conflict or abuse against workers.

“But what if the workers designed a textile pattern and pattern themselves, rather than the common patterns and patterns that have been passed down for many generations.

“If a factory that uses mechanized textiles uses the same pattern without the permission of the workers, it is an infringement prohibited by law.

“The objects of machine learning happen to be creative products, and most of them are copyrighted products.

“So machine learning should be licensed by the copyright owner of the learning target.”

After listening to this, Zhu Jingyuan looked at Judy Pu:

"Boss, do you agree with this reason?"

Judy Plutonium said directly:

"Brother Lan... your statement makes no sense, and the metaphor of machine learning and textile workers designing patterns is also quibbling.

"The process of machine learning and the subsequent re-creation process are not essentially different from human learning." When people learn other people's works, they essentially break the elements of the work into pieces and store them in the brain in a molecular or even atomic state. in.

“Then we combine these atomic materials by ourselves to create a work that looks completely different from the original work.

“It’s just that these atoms are too small to identify direct similarities to the original work.

“In fact, it is just based on the rules for human beings to identify plagiarism, and continues to be subdivided infinitely until humans cannot recognize it.

“Are the same design patterns and paragraphs considered plagiarism? Are the same character structures and outlines considered plagiarism?

“Does copying the same design elements and verses count as plagiarism? Does the same line style and phrases count as plagiarism?
“Does the same color range combination and vocabulary range count as plagiarism?
“Continuing this breakdown, humans learn to end up using a similar range of colors and vocabularies.

“The depth of machine learning is probably much deeper and more granular than this.

“If you use human legal plagiarism identification standards to identify works created by machine learning, you will never be able to identify plagiarism.

“The analogy between hand weavers and mechanized textile machines can only be regarded as using similar cotton and colors.

“So why is it not plagiarism when humans learn, but plagiarism when a machine does the same thing?

Judy Lan smiled and said:
"Second brother, you may have misunderstood what I meant.

“I didn’t say that machine learning and creation are plagiarism, but human learning and creation are not plagiarism.

“Actually, according to my own point of view, both should be considered plagiarism.

"As you said, second brother, as long as the standards for identifying plagiarism are infinitely deep, all learning can be judged as plagiarism.

“On this basis, anyone who has studied other people’s art for free should not create any work for profit.

“If I want to create works for profit, all my learning materials should be materials licensed for profit.

“Because as long as you have studied, you will obviously be affected to some extent, otherwise it cannot be called learning.

“The criterion for judging whether it is plagiarism or not is, more accurately, the level of plagiarism allowed.

"Second brother can refer to the logic of calculus and relativity to understand this statement.

“On this basis, real human society and human creators have, through actual social practices and laws, tacitly authorized other humans to have a profit-making license for plagiarism below a specific level.

“However, humans have not empowered machines to use any level of plagiarism monetization standards.

"For example, if I draw a painting, I will allow you, my second brother, to learn how to paint by default.

"I allow you to store my drawings of atomic structures and even molecules in your mind and use them at any time.

“But you are not allowed to store them in the machine for use by the machine.

“In order for a machine to learn, it also needs authorization.

“Because you said that the logic of machine learning is the same as that of humans, but machines happen to not be humans.

“There are no existing laws related to machine learning.

“Machine learning has not yet appeared on a large scale, and I have not directly authorized it, let alone factual authorization.

“You’re not even given permission to use machine learning.

“Now we have to draw up laws to determine how much machine learning should be granted by default.

“My suggestion is that everything is non-default, and any material used for machine learning must be directly authorized by the creator.

“The purpose of this is not to prevent machine learning from continuing to evolve.

“It’s precisely about regulating the machine learning industry.

“The future is the era of big data, and the scale of machine learning will inevitably become larger and larger.

“If the scope and rules of machine learning are not limited, it will inevitably lead to the chaotic development of this industry.

“It will inevitably lead to machine learning developers plagiarizing each other and being uncontrollable.

“Of course plagiarizing data atoms is not plagiarism, then plagiarizing databases must be plagiarism, right?

“Just like copying the scope of a dictionary is not plagiarism, but copying a large dictionary is definitely plagiarism.

“In the new era of big data and machine learning, data materials should have new ownership specifications.

“Any data and information should be owned and valuable.

“I think this should be the logic of the times in the next era.

“With such basic rules, there are basic rules for standardized charging and authorization.

“Only the corresponding data model developers will have the opportunity to use authorized data to develop equally authorized data models.

“Then you can use the products produced by the machine to make a profit without additional restrictions.

“Otherwise, the copyright ownership of any work ultimately produced by machine learning will be vague and sensitive.

“Machines are not human beings, so humans do not need to protect the copyright of machines and do not recognize the copyright of works created by machines at all.

“Only if the machine is regarded as a human machine, the copyright of the works created belongs to its owner.

“Only by using the logic of human copyright can we obtain protection and profit.

“But if humans want to make money from other people’s materials, they must obtain the other party’s authorization, which will form a clear copyright ownership.

“If you look at the essence through the phenomenon, using mechanical intelligence to complete tasks that only humans could do before, and increasing the output and speed by dozens or hundreds of times, is another industrial revolution.

“Productivity will increase by leaps and bounds again, the production costs of products covered by technology will drop significantly, and the resulting benefits will increase significantly.

“Those who provide the raw materials for the successful operation of smart machines have the right to share in the benefits brought by this increase in productivity.

“The license for personal appreciation of learning cannot be used to develop and manufacture any production tools.

“In this way, the logic of data valorization will be clear.”

After listening to this, Zhu Jingyuan looked at Judy Pu:

"Second, what do you think?"

Judy thought silently for a few seconds, and finally nodded slowly and said:

"I think what Di Lan said makes sense...

“It is a very good idea to use the logic of calculus to judge and distinguish the boundaries of learning, imitation, and plagiarism.

“On the future public network platform, when creators upload their works.

“There should be an option to allow machine learning or not.

“There should also be an option whether to use big data to filter and recommend information.”

After hearing this, Judy Lan immediately agreed and said:

"I agree with the second brother's opinion."

Zhu Jingyuan smiled and said:
“Since you two brothers think it’s feasible, let’s let these things be settled.

“But the question is, how do you determine whether a data model was trained using unauthorized materials?
"Do you have any way to distinguish them at the data analysis level?"

Judy Plutonium thought for a while and said:
“Some validation data can be added to unauthorized footage, circumventing the use of unauthorized machine learning or even polluting data models.

"However... it is not possible to rely solely on technical means, and it is impossible to completely circumvent the cracking and screening of violators and the anti-sabotage technology.

"So we still need to use social and institutional restrictions and post-event disposal methods..."

Judy Plutonium said this and looked at her brother.

Judy Lan thought for a while and said:
“There are two directions in the system. One is to default to all authorizations and at the same time default to all charges.

“Currently, the large-scale art platforms on the Internet are official platforms of the imperial court and the four major industrial groups.

“Private manufacturers and users who want to build their own data models can only collect data from these platforms.

“So it can be stipulated that manufacturers who have not paid one of the existing platforms to obtain authorization qualifications will be directly punished for violations.

“Another direction is for the four major industrial groups to come forward and design a basic software framework each.

“It is equivalent to a device with forced filtering and filtering functions.

“Any data training conducted by private manufacturers must be developed based on this basic software framework.

“That is, put it inside this filtering device.

“This filtering process automatically identifies unauthorized material flags and directly excludes unauthorized material from the training process.

“At the same time, there should be special markings on authorized and unauthorized material on the Internet.

"For direct identification of these filtering devices.

"These plans can be implemented simultaneously with the technical plans of the second brother."

Judy Pu could not help but nod after hearing this:

"Well, identification and counter-training techniques have to be developed."

Zhu Jingyuan did not comment on the two people's ideas, so he asked Zhu Zhu Pu and Zhu Zhu Lan to discuss the matter.

(End of this chapter)