The Lion City seems to be an independent country, and to some extent it is still dependent on Amerikan for its existence.
The official website of Amerikan writes roughly like this:
The two sides have established an extensive and lasting relationship based on common economic interests, defense and security cooperation, and lasting people-to-people relations.
In recent years, Amerikan has actually been strengthening diplomatic relations with the Lion City.
Among them, a fact sheet on strengthening the strategic partnership with the Lion City was released in 2021. This was repeatedly emphasized during Vice President Amerikan’s trip to the Lion City that year.
From 2021 to 2031, the relationship between the two parties has been strengthening.
The Li family changed hands and sold the Lion City to Zheng Li. This was something Amerikan did not expect. By the time they found out, the Li family had already completed intention negotiations with Zheng Li.
The intention is over, and it's too late to stop it now.
And for Amerikan, the Lion City is not that important compared to the benefits that may be obtained from Zheng Li in the future.
So they remained on the sidelines of the deal.
And the Li family can make deals with Zheng Li, but Amerikan can't?
Under the control of the truly powerful Amerikan family, Zheng Li obtained all of Amerikan's men in the Lion City with three potions that could extend his life for ten years.
At the same time, the previously signed agreement on leasing a military base in the Lion City was abolished. Since then, Amerikan’s interests and organizations in the Lion City have completely withdrawn.
vertex
The Lion City councilors who originally thought they could pull off a tiger skin, that is, Amerikan, lost their temper in an instant.
Amerikan has honestly withdrawn its troops, so why bother?
As for contacting the armed forces of the Lion City to carry out a coup, this is pure wishful thinking.
Merlin's image to the outside world is not an entity at all. At the same time, he takes over the artificial intelligence of the Lion City. How can your armed forces rely on to kill the artificial intelligence?
Furthermore, this is the result of a national decision. You have no legal basis, and there is no substantive physical goal at all.
In addition, the armed forces of the Lion City are not stupid. They are gambling their own future or even their lives for the benefit of politicians?
The remaining more than 100,000 civil servants in the Lion City were divided by Zheng Li.
Zheng Li used the name of the Lion City to establish several new state-owned enterprises.
These companies will recruit from a broad pool of civil servants.
If you are honest and obedient, you can participate in the selection. If you are dishonest and obedient, you can go ahead and make trouble. Zheng Li expressed his attitude that it was impossible to compromise.
At the same time, the benefits of the unemployed will be greatly increased, and money will be paid directly according to the minimum guarantee.
After the combination of punches, all the unemployed civil servants became obedient.
Everyone knew that Merlin would not do any behind-the-scenes operations. Since it was a selection examination, these civil servants who selected the best among the best naturally felt that they had a chance.
And for them, the future of the state-owned company established by Merlin is bright, both in terms of treatment and the technology embodied by Merlin.
During the intervening period, Lion City released a series of artificial intelligence frameworks, writing in its framework:
“The artificial intelligence framework announced below follows the open source spirit that has been popular since the rise of Blue Star Internet and makes public some AI frameworks that are better in some aspects than Blue Star’s existing AI framework.
All individuals and enterprises can use these AI frameworks for free. "
In Zheng Li's view, Blue Star's productivity is not improving fast enough, and the progress in integrating artificial intelligence with other industries is still not as expected.
After the government-oriented artificial intelligence in Lion City fully demonstrated his muscles in the field of artificial intelligence, he hopes to further promote Blue Star's artificial intelligence and industry integration progress by making the AI framework public.
He can complete the entire process, or he can bring some inspiration to Blue Star's artificial intelligence field and let them complete it under his own guidance.
The further implementation of artificial intelligence, combined with biophotovoltaic technology, will reduce Blue Star’s employment population by 80% or even more.
A large amount of productivity has been released. Whether it is to engage in more technical work or to enrich the virtual world ecology, it is better than doing some meaningless work now.
Zheng Li's stay at Blue Star is not short. Although Zheng Li has been at the top of society for more than ten years, through the Internet, he has a certain understanding of Blue Star's social ecology.
In his opinion, many jobs at Blue Star are meaningless and exist purely for the purpose of torturing people, without even producing any output.
The manufacturing and service industries are at least producing output.
Even if it is productive work, Zheng Li still feels that it is meaningless. The intelligent life that was finally born in the vast universe should not spend its life in such boring mechanical labor.
Intelligent life needs to be allowed to find the meaning of life on its own.
Rather than considering the many levels designed by society, even if you stumble through all the levels, you will still feel like your life is wasted.
For Zheng Li, the Lion City is just a testing ground, and it will eventually be promoted to the entire Blue Star.
The AI framework disclosed by Zheng Li under the domain name magi includes a new programming language and open source packages surrounding this programming language, that is, tools surrounding this programming language.
The reason why python has become so popular in recent years is not unrelated to the fact that python has a large number of useful toolkits.
These encapsulated packages greatly reduce the difficulty of programming in python language.
As soon as the M language released by Merlin was announced, it quickly caused an earthquake in the entire Internet industry.
"Too humble, not only excellent in some aspects, but excellent in all aspects.
Although I haven’t understood it yet, I always feel that this thing is much more advanced than Blue Star’s current mainstream artificial intelligence framework. "
The main reason is that this thing is too professional, and anything that is too professional is destined to be insulated from the mainstream group.
Even professional algorithm engineers need a period of study.
Most of the comments from netizens were concerned about whether they would lose their jobs:
“It feels like Merlin is not a magician, but a programmer, trying every means to promote the large-scale implementation of artificial intelligence.
First, artificial intelligence was used to replace human labor in the Lion City, and then open source AI programming languages and frameworks were released.
I seriously doubt that magicians use magic a bit like programming? "
"Merlin should have taken away his magical powers. It was hard for me to find a job. Why don't they optimize me again? I don't want to graduate from Weibo.
This is the ideal job that I found with great difficulty. "
The person who posted the following Weibo post is a yellow connoisseur.
Such careers do exist, and they pay well.
But for him, it was a prophetic word.
After Merlin released the AI language and supporting frameworks and toolkits, these major Internet companies asked the following algorithm engineers to put down all their work:
“Everyone’s only job during this period is to learn how to use the M language, M framework and supporting series of packages.
Then combine our work to see which places can be rewritten using M language. "
Among them, the first one to be used was AI pornographic identification.
AI anti-pornography has always been a core requirement for content security. Basically, everyone from Pig Factory and Goose Factory to Weibo and Byte are studying this stuff.
It is a classic that is easy to get started but difficult to master.
Early yellow identification was basically manual review, which was a labor-intensive type of work.
Later, as more people went online and there was more content, the cost of manual review became higher and higher, so the use of AI + manual methods to identify pornography became mainstream.
The AI+manual method generally uses a machine to filter out most of the images that must be normal and must have problems, and then hand over the rest to manual review. This can greatly reduce labor costs, and the better the machine recognition effect, the lower the cost of manual review. The lower.
AI pornographic identification is actually a relatively broad concept, which can be implemented through a rule system, such as setting up a blacklist database based on MD5, user IP and other information, and intercepting directly based on rules.
Most of them still use algorithmic models, that is, they use algorithmic models to determine whether an image contains sq information, which is essentially image recognition.
Image recognition is currently even better than humans on some tasks.
The most common image recognition algorithm is image classification algorithm, from Alex to VGG, from Res.
The current image classification algorithm can more accurately distinguish the 1000 categories of Image data. Pornographic identification itself also classifies the input image, so it is natural to adopt the image classification algorithm.
Moreover, the target detection algorithm can be used to detect dew points in sq images, which is also a relatively reliable method.
In addition, there are features and logic based on business-level structures, such as whether there is a person, the area of the skin, etc., to assist judgment, which is indeed effective in some cases.
The difficulty of identifying pornography with AI mainly lies in non-revealing soft SQ, SQ with small features, non-universal SQ, cartoon and animation SQ, etc.
This is the difficulty in identifying pornographic images using AI, and it is even more difficult in identifying pornographic images through video and audio.
And for these major Internet companies, even if they can achieve a 99% interception rate, the remaining 1% of the content will not matter.
Taking Weibo as an example, the data generated every day is measured in T units.
Even if there are dozens of Ts, a 1% interception error rate is enough to make the whole situation miserable.
What's more important is that China's content review industry is not limited to pornography detection. Major Internet companies such as OCR review have been doing it for a long time.
Click till then.
For the same reason, an interception rate of 99% is unacceptable to them.
So until 2031, the method of identifying pornography will still be AI plus manual work.
It’s just that in 2021, a company the size of Weibo may need thousands of yellow appraisers, but by 2031, it will only need hundreds.
Four digits become three digits.
As for WeChat, you do it once and then throw it to the back-end review interface.
A weighting algorithm is used to determine whether you have violated the rules. After a specific rule is triggered, the weight of your tendency will be determined. After the weight exceeds the threshold, you will be focused on.
There will be a manual review for you.
Of course, this kind of review is not limited to major domestic manufacturers. Facebook, Instagram, YouTube, Google and Twitter all have a large number of manual review teams.
Their review team is based in the Philippines.
In 2018, PBS released a documentary on this matter.
For domestic and foreign Internet giants, they can achieve close to 100% in text recognition, but in image classification tests, they can only achieve about 98%.
Moreover, it has very high requirements on computing power and cannot be used in actual production environment at all.
This is the result of Image’s annual image classification test competition. Image and video recognition in actual operations is much more difficult than the Image competition.
After Zheng Li disclosed the AI algorithm, he used the model written in M language.
From deployment to use, it has broken through the cognition of these Internet companies. An AI model that can achieve a 99.9 recognition success rate for content requires about the same computing power as before.
In other words, the scale of AI pornography detection by hundreds of people in major Internet companies has now been reduced by 90%.
AI pornographic identification is just the first application of M language.
The AI technology it represents has greatly improved its computing power advantages and the difficulty of implementing the algorithm itself.
It allows for the easy construction of most otherwise complex architectures and can be widely used in industrial fields.
A large number of modular components that are easy to combine, writing your own layer types, computational graph abstraction, data and model parallelism, dynamic type stability, etc. are all perfectly adapted.
After algorithm engineers used it, M language-related projects on Github became popular in the community.
In foreign programmer communities, active experts are complaining, and some experts who rarely speak are also making comments:
"Merlin's technology is far beyond our imagination. If he developed the M language himself, his achievements in the field of AI would exceed all of us.
It is really rare to see an AI programming language without any flaws.
For example, Caffe can port Matlab's fast convolutional network implementation to C and C++, which is suitable for feedforward networks and image processing, and the model can be trained without any code.
However, caffe cannot be used in recurrent networks, and its scalability is extremely poor.
The TensorFlow framework launched by Google can generate calculation graphs and then perform automatic differentiation, eliminating the need to manually code when trying new neural network arrangements.
But it runs very slowly and is very prone to errors in large software projects.
Basically, the AI frameworks launched by major Internet giants in recent years have more or less problems.
Of course this is unavoidable and nothing is perfect.
Just like how many people use Java and no matter how long they have been active, they can still find shortcomings.
But M language seems to have no shortcomings in the field of artificial intelligence.
It seems that after countless tests and optimizations, it has reached a kind of perfection.
This makes me seriously doubt that the artificial intelligence Merlin masters is strong artificial intelligence.
He can realize almost all the fantasies of Blue Star people about artificial intelligence.
Including self-programming.
It doesn't seem that strange that one piece of code can compile another piece of code.
After all, current artificial intelligence models can already be programmed simply by themselves. "
