bet365 Play online games_not worthy of it as a

Home>Journal of China Social Sciences>Special plan>Special planning content

Calculation creativity and calculation criticism

—— A short discussion of a kind of AI criticism is possible

00:58, April 03, 2020 Source: "China Social Sciences", April 3, 2020, 1900th author: Zhao Wei

　　In the "Big Data" era，Artificial Intelligence as the core technology to realize a strong technology country，Not only becoming a country’s strategic needs，It has also led quantitative humanistic research in traditional fields。What does artificial intelligence mean for humanities，How should humanists use bet365 Play online games the relationship between quantitative methods and digital technology，These are related issues that need to be considered from the perspective of digital humanities。

　　Artificial intelligence in digital humanistic practice

　　Different theoretical thinking and ethical worries in the philosophy of theory，For digital humanistic practice，Artificial intelligence has been more or less for a long time to mean the practical field of machine learning (Machine Learning)，Especially the unsuccessful learning (UNSUPERVISED Learning)，That is, the data training method with almost zero manual intervention。Here，Artificial intelligence is not the "Super Smart" of God in science fiction novels，More statistical methods at the end of the tentacle，It is becoming more and more commonly used in large -scale data modeling，Related practitioners have also been learning、Experiment in learning。

　　From August 2009，The University of Chicago's Text Optical Laboratory uses the Republic of China periodical periodical corpus data provided by the Shanghai Library，Orthodisculated from five directions to distinguish and discover the methods of various potential "new literature" in 1918 to 1949，Among them, there are at least three paths involving deep learning，SEQUENTIAL MODELING and Pattern Recognition technology。Apply some practices in the industry to literary research with high accuracy requirements，Not only will it not make people feel dazzling，On the contrary，From experimental design to program operation、Then to the necessary statistical test，Unlimitedly full of all kinds of failures and unsatisfactory。In order to achieve a certain accuracy，We must continue to give up some seemingly Bet365 app download correct practices，Keep trial and error and "compromise"。In fact，It seems that rarely thinks that this is artificial intelligence，Maybe the digital industry that is constantly active will be more interested in such labels。

　　About artificial intelligence applications，Another phenomenon that the literary and artistic world is keen to talk about is AI writing。In this regard，Sun Maosong team of Natural Language Treatment and Social Humanities Computing Center of Tsinghua University has been doing related explorations。The "Nine Songs" poetry machine from the team's laboratory can not only make five -character seven -character quatrains，Can also be used as a rhythm，I also participated in the competition show many times。Sun Maosong proposed the concept of "Calculating Calculation" in the "Digital Humanities" created in December 2019。But can the machine be "created"? not necessarily。Because all this is possible，It is nothing more than statistical。From Sun Maosong's view，No matter how complicated deep learning, it is based on a possibility for infer。Taking only seven laws as an example，If roughly assume that the universal Chinese characters used in Chinese classical poetry are about 10,000 words，So theoretically the total number of poetry (arrangement combination) that can be produced in theory will be 744.。This is an astronomical number，So you can only rely on algorithms to dig out "good poems"。How to dig? You need to model，It is necessary to use more than 800,000 lines of ancient poems and the labeled knowledge base as the material，Come to train a model designed for specially designed。These models use the information circulation processing process similar to "work memory" in cognitive psychology，The possibilities of various vocabulary in the poetry bank have been arranged one after another。In this process of constant "cache"，The theme of the poem noted by the keywords and the historical and cultural information provided by the relevant corpus are stored in a dynamic model，and the setting of the corresponding algorithm has greatly guaranteed the continuity of the connotation of the poem line，Finally supplemented by the study of rhythm rules and style patterns。A large number of user score experiments proves，bet365 Play online games "Nine Songs" can not only make peace、Coherence、Poetry of "Artistic Contents"，You can also cooperate with people to write poems，Poetry scoring for autonomy。That is to say，When AI "Poetry"，Models and algorithms still come from people，From the "acquisition" from a large number of ancient poetry masterpieces，These templates are reflected in the rules level generated by syntax。

　　Theoretically speaking，The model after training completes a "integrated" work。So a critic said，Poems written by AI as "industry access" principle，Not as much as they write, not worthy of it as a poet。Of course it is a joke，But from the current mechanism of most machine poetry at this stage，But it doesn't make sense，Because "standing on the shoulders of the giant" is certainly worse than the predecessor。Just as a literary researcher，We are not concerned about whether the "nine songs" can make a good sentence beyond the classics of ancient people，We are more concerned about the mechanism behind it by automatic poetry，What are the essence of the grid。Today，For example, how to use the algorithm of the exquisite experimental design to indicate the feeling pattern (Pattern) and a probability factor that determines the "pattern" of ancient poetry、What is bet365 Play online games the relationship between the external rhythm and the inner music，The topic of digital humanities is gradually becoming a digital humanities。

　　Algorithm black box and calculation criticism

　　Digital humanities are to reasonably introduce the use of calculation tools into proof of humanities research、Inference、Analysis、Explanation and even criticism，This is by no means a discipline that can break through one by one，It must be a valuable problem in the collision of interdisciplinary，Then solve it through teamwork，Finally obtained the internal or external recognition update of the field。Digital humanities can of course be very dazzling，can also be very flat，But the cross -disciplinary cooperation that really breaks through the border is extremely difficult，It is not more difficult than developing an artificial intelligence as a poetry machine。These difficulties not only come from technical operation，More from the evaluation mechanism，and some methodology and conceptual prerequisites that must bet365 live casino games be fully discussed。However, the problem now is，The technical hotspots driven by artificial intelligence are endless，What caused the data driver with the trend、Model driver or knowledge or problem -driven? During this process，Where is the position of humanities? Or，How can humanist scholars avoid becoming a vassal of technology?

　　Starting from these problems，The author believes that the scholar of digital humanities or cultural analysis (Cultural Analytics)，It is possible to reflect the value care of humanism through its own criticism practice。This is first reflected in the dare to achieve a "removing charm" for technology，This is precisely a kind of effort to improve the scientific and technological literacy of humanities，or it provides a cut -in port for it。This is not only because "in it" can better "beyond"，It is also what literary researchers can do。To put it simply to calculate criticism，3956_3972。Current，Artificial intelligence is likely to cause people's dissatisfaction. The so -called black box。Talking about a data science application research, you can't help but talk about algorithms，Just like understanding a poem that cannot cross its language。However, many deep learning based on big data，The complexity of the calculation process is "beyond the scope of human understanding"。For example，If you want to study the evolution of the narrative network in a novel in a certain period，May first need to extract bet365 Play online games the relationship between the characters in the characters in all texts。When the number of text is small，It is limited to the scope of a single writer，You can observe hundreds of thousands of times，Design or find a series of rules according to the characteristics of the situation，To determine the dialogue character，Then use the statistical tool to give the entire possibility relationship in the text to "calculate"。This is a typical top -down、A clear approach at a glance。but，Once the number of text rises，Extension to thousands of novels，It is necessary to invent an automatic recognition tool for the character。If you do not combine artificial labeling at this time，Simple supervision and learning is unspepped。Even if it is based on a certain manual labeling and given rules，It may also be extracted all the relationship between the sentences，As for what the computer is based on what algorithm is done，No one knows。The relationship extracted from the big data driver，If it is not appointed, it is directly used for the next literary interpretation，It will be difficult to form an interpretation meaning，This is also a large number of problems that are available based on "common" technology。This kind of bottom -up、The practice of less intervention is difficult to cause people to question。So just as Sun Maosong said，Black box may be a good thing for technical exploration，Turing test is a black box。This is because the technical breakthrough is more concerned about whether it can achieve the purpose，Successful realization of a certain path，However, the literary research that discovers and interprets equality here，It may be meaningless。

　　How to carry out and its meaning

　　So，At present, at the current high -precision knowledge driver，Do you want to continue the exploration of digital humanities? The answer is yes，But you need to consider how to join the "person" link at this time。The so -called calculation criticism，A word to cover it，The result of "people read"，Come to question the results of "machine reading"，Through experienced comparative analysis，Connecting the deeper level of the possible cultural causes。

　　Research on English Sentences in the Modernist period in the period of modernist in the United States (Hotet Lang、Su Zhen: "Literary Mode Identification: Modernism between Text Reading and Machine Learning") as an example。The technical path of this research is relatively ordinary，is to let the machine society judge whether any poem text is "俳" or "non -俳 sentence"。But for two literary researchers，The real experiment begins to start with the part of the "misjudgment" of the machine，To figure out the statistical model used to identify the text mode of 俳俳俳，How to coordinate with the interpretation of the body view of the body based on the text based on text and the historical and cultural interpretation。That is to say，Once the machine is wrong，errors have triggered an important "interpretation problem": what makes certain poems (or certain features) such "like a sentence"? Check the surface of Bet365 lotto review a poem，They found，The original algorithm was determined based on the probability of several specific words and syllable characteristics -to some extent, "cracked" the black box of the algorithm。The meaning of cracking is of course not to explain the limitations of the machine，Instead, these "potential sentences" created by machines，Just created a new thing for literary research。It is half of the program script，The other half comes from the history of literary criticism，This not only gives the algorithm a place，and make the language characteristics at the individual level that may be scattered on the level of hundreds of poems of dozens of journals，It reminds "A broader text collection shared with a specific element of a sentence"。This collection will also include the translation and adaptation of the characteristic mode in the adaptation，provided "both related to the style of the sentence、At the same time that belongs to some extensive oriental atmospheres "，This atmosphere is "the text mode of the circulating" that people imagine。This kind of remote reading show，Researchers can examine the social and historical influence and status of text in the broader poetic discourse，Then pursue the cultural logic of the hidden American modernist period behind the sentence。

　　In fact，This quantitative literary criticism，or Calculation Critic (Computational Criticism)、Algorithm Criticism's statistical basis is a model of "self -resistance" (Su Zhen)，The essence of this approach is the interaction between people and algorithms，It is a kind of emphasis on people's consciousness and machine dialogue、Cultural criticism of confrontation。In a criticism of a researcher，On the one hand, the subject completed the questioning and test of the algorithm，On the other hand, the algorithm also liberated the criticism potential limited in the original field of literature，The intermediary concept generated by the algorithm，People can return between remote and careful reading，constantly approach the problem you want to study。

　　The starting point of advocating calculation criticism is also to be used to oppose digital humanistic tools。Essentially like Ted Underwood，Every time we use a search engine search information on bet365 live casino games the Internet，In fact, I have completed an interaction with the algorithm。"Humanists" usually tend to ignore the rich meaning of this action，However。It can be said，That kind of background technology、The theory of theory、The Returning of the Polytechnic、Humanities of humanities，It is still an identity that is not observed by the current tools of tools.。The position of humanism should not be the key obstacle of people's use of quantitative means to access the problem。On the contrary，Real humanists must even restore the "all -talented" state，This is the meaning of the development of digital humanities。So if we are too clearly drawn digital technology as a device in shape，Respect traditional learning as a way to shape，Maybe the color of the tool theory may be more important。So push it away，At this stage，A landing、Artificial Intelligence Criticism in the cut，Can't nor around the algorithm。

　　(Author Unit: Institute of Literature of the Chinese Academy of Social Sciences)

Editor in charge: Changchang