Author | Wang Hao
Cultural studies is a humanities discipline that has been developed for hundreds of years. However, cultural studies have always been a research discipline that uses small-scale data due to limitations in data scale and geographical area. With the advent of the big data era, public data sets such as Internet user behavior data have become the latest gold mine in the field of humanities due to their large data volume and rich information. In 2022, researchers published an article introducing computational cultural research at the international academic conference MHEHD 2022, describing how to conduct cultural research through artificial intelligence algorithms.
This paper mainly analyzes the sociological effects of a zero-sample machine learning algorithm called ZeroMat. The ZeroMat algorithm is the first algorithm in the field of artificial intelligence that truly does not use any data for recommendation. As we all know, existing zero-shot learning algorithms are basically variations of transfer learning and meta-learning. And ZeroMat is the first algorithm that is different.
The ZeroMat algorithm assumes that the user item rating matrix obeys the following distribution:
##Put the above formula into the probability matrix distribution model and get the following formula:
After taking the logarithm, use stochastic gradient descent to calculate U and V, and set the variance to a constant 1, the following formula is obtained:
From the formula We found that this algorithm is essentially a zero-shot learning algorithm. The sociological effect of this algorithm is that our culture has been locked in, because we can very accurately predict the user rating data of cultural consumer goods such as movies without any data, that is, we can know the user's personal cultural preferences without using historical data. And because of the high imbalance of user rating data, we know that human culture is not only locked, but also locked in an extremely unequal extreme state, and it only takes a short time for all of this to happen.
In fact, this phenomenon not only exists in the cultural field, but also widely exists in other social fields. Many measures taken by the Chinese government have alleviated this problem to a certain extent. For example, the big city cluster strategy utilizes the principle that the Matthew effect helps improve efficiency and promotes economic development. Another example is the government's supervision of Internet celebrities, which effectively reduces the Matthew effect of information dissemination and prevents information dissemination from falling into a locked state.
The fundamental reason why human culture is locked up lies in the effect of power law distribution in various social phenomena. The reason why power law distribution exists is because it can maximize efficiency and is related to diversity. Seriously studying power law distribution can allow us to better understand various social phenomena, and help us better study algorithms and other natural sciences.
Is Human Culture Locked by Evolution : https://www.atlantis-press.com/article/125975737.pdf
Introduction to the author
Wang Hao, an Internet veteran and former head of Qujia Games AI Lab, has more than 11 years of technology and technical management experience in Internet companies such as Douban, Sina, and NetEase, and has successfully launched more than 10 technology products. He has rich experience and unique insights in the fields of recommendation systems, risk control and natural language processing. Published 30 papers in international academic conferences and journals, and won the Best Paper Award/Best Paper Report Award at international academic conferences three times. 2006 ACM/ICPC International Collegiate Programming Competition North American Rocky Mountain Regional Competition Gold Medal.
The above is the detailed content of Is human culture locked in evolution?. For more information, please follow other related articles on the PHP Chinese website!
