The printing industry is moving towards intelligence and automation
In the process of digitizing printing technology, many experiences of using computers to help manufacture have been accumulated. If these experiences are applied to the management system, help the computer to judge, and start the next process, this is already smart printing.
Recently, a friend needs a description of a foreign product. He only has the English version of the instructions on his hand and posted it to Google's automatic translation. The whole two-page technical description, after being translated into Chinese, does not use a word. The changes can be transferred to the customer.
At the end of last year, the Google translation algorithm was replaced by the Nueural Machine Translation (NMT), which replaced the original NLP natural language processing. The translation quality was almost as good as the experience of experienced technical editors.
The translation machine adopts the most popular "machine learning" technology of artificial intelligence, allowing the machine to learn two languages like a human being. The machine is not sleepy and the calculation speed is fast, and it can quickly progress from the primary level to the expert level. The speed of improvement is fast.
What is machine learning? It is to organize the processing of things into logic that can be calculated by computer, and then provide enough samples to let the computer practice and compare the answers according to the algorithm. Just like when we were young, we had enough poetry, and some people were enough to read and immersed in a large number of poems. Among them, the logic of finding poetry can finally make poetry right, satisfy the basic elements of poetry, and make amazing works.
Machine learning In today's environment where computing power is easy to get, teach computer learning at night, we feed the known poems, and the computer can already write poems tomorrow morning. Even if the computer writes poetry not to express emotions, it can also touch people, because the source of data he learns is originally the source of human emotions.
Machine learning pushes "artificial intelligence" to the computer to learn any work done by humans. As long as the "content" of the work can be depicted by "digital", the computer can learn, whether it is esoteric Go or a simple money counter. ...
Machine learning is not a new thing that is available today. It is just that the methods and environment of machine learning are more efficient today. The digitization of all walks of life has always been the process of professional teachers teaching computer engineers and engineers teaching computer learning.
The printing industry is the best example. From the CDR files to the prints on the plates, to the ink keys of the ink control station, to the delivery location of the courier, if you use a computer to help, it must be more efficient than people. This is because for more than a decade, software engineers have learned the content of each process, using numbers to characterize the printing process, and finally converting it into computer software that helps with print production and management.
Almost all processes in the printing process have already been digitized, and what can computers learn?
As long as the parts that need to be handled by humans are worthy of computer learning, technology suppliers are aiming at solving the key to labor costs in order to help printer customers improve their competitiveness.
Automation has replaced many people, and what parts need to take automation a step further?
Today, the printing house hired a lot of PS masters to process the files, because the files sent by the customers are not standardized. There are several layouts on one file, which must be sorted to be converted into PDFs associated with the orders. If the computer learns to be able to judge the extent and front and back of the layout like a human, and automatically process and output the PDF, the bottleneck of organizing the files can be opened. This is the first item that the computer needs to learn.
The computer has already managed the order, the managed order is printed as a construction order to direct the production process, and the second is how to teach the computer to change the content of the construction order into a command to direct the prepress process. After the order is confirmed, the production system Automatically started working spontaneously.
There is still one person who can't solve it. The cost can't be controlled is the loss caused by the error in the file conversion. The prepress process is from the design manuscript, the PDF version is exported, the small version is made into the large version of the PDF, and the RIP is refined into the CTP version. The print on the top, and finally on the machine, each of the above procedures re-combined the file content, and then handed it over to the next process. The PDF file itself is the programming language, and the surface text or image is occasionally changed during the combination process. The attributes of the layout, the next step, the layout content has been changed.
This kind of mistake can only blame the computer, and it is afraid that if the error appears on an important print, the cost cannot be estimated. Since different processes use different vendors' software, it is impossible to teach computers to make mistakes. The solution is to teach computers to review the manuscripts, take out the files at different stages between the processes, and check them out before CTP plate making. The layout.
At present, different domestic technology suppliers are working hard on these three aspects. Machine learning requires the same experience as human learning. The more experience, the stronger the computer's execution ability.
The ATMs at the high-speed rail station provide ticketing and ticketing services. The check-in tickets are also replaced by machines. There are similar machines at the station 20 years ago, but today's machines are smarter because of their experience. Engineers learn to improve the function of the machine, or the machine will learn by itself, all of which can lead to intelligent manufacturing.
Today, machine learning technology has become the main driving force for the rapid advancement of artificial intelligence. If we think that artificial intelligence will only happen in a specific field and has nothing to do with the printing industry, it is the ignorance of today's technological development.
At present, the pursuit of transformation of printing management is to teach computers to learn to organize documents, direct production processes and automatic review, and until the various automation applications mature, smart printing will take a big step forward.