https://news.cnstock.com/news,yw-202307-5091780.htm
The emergence of ChatGPT has attracted great attention from all walks of life to large AI models. The emergence of large AI models is not accidental. It is the product of long-term theoretical research and application innovation iterations. For most non-professionals, large models are a relatively unfamiliar professional term and technical concept. In the book "The Era of Large Models: ChatGPT Starts the Wave of General Artificial Intelligence", two professionals discuss the technical application and industry of AI large models. Changes are analyzed in depth. Although this book spends a lot of space focusing on the specific applications of AI large models in education, law, human resources, corporate services, creative entertainment and other fields, the author's discussion of the technological evolution history of AI large models and the development history of the three industrial revolutions is evident. The depth of research, especially in the understanding of the relationship between technological innovation and human social progress, has gone beyond thinking at the technical level and has more social and humanistic attributes.
Theoretical Evolution of AI Large Models
To understand the large AI model, you first need to clarify the concept and definition of this technology. According to the definition given by the author in this book, AI large models are based on algorithms such as mathematics and statistics to describe a system or a data set, so that machine learning and model training can be more accurate and effective. function mapping. The training and adjustment of very large models require extremely huge computing resources and data volumes. The reason why the application of artificial intelligence explodes is mainly due to the gradual maturity of conditions in two aspects: First, the AI large model technology is becoming increasingly mature, the underlying technology is developing rapidly, and research on language processing and neural network architecture has made greater breakthroughs; second, AI The application of large model technology has entered the market and the public. Technology companies represented by Google, Meta, and Amazon have invested a lot of resources in the development and testing of large models and developed a large number of application markets.
Although large AI models have only become popular in the past few years, in the professional field, the theoretical research on artificial intelligence technology has lasted for nearly 70 years. According to the author's introduction in the book, artificial intelligence first originated from the Artificial Intelligence Conference at Dartmouth College in 1956, which then led to three basic schools of artificial intelligence: semiotic school (logicism, computer school), connection school (bionic school) science), behavioral school (evolution or cybernetics). From the early semiotic school to the later rise of computer technologies such as bionics and evolution, AI has experienced three milestone stages of development, namely machine learning, deep learning and artificial intelligence content generation large models, and the popular ChatGPT is a large model Typical of the stunning effects produced. According to the author's explanation in the book, the training of large models requires a large amount of computer resources and data, so it integrates many of the most cutting-edge computer technologies and becomes the master that triggers the AI revolution.
Since large AI models involve a lot of professional knowledge and technical terms, it is difficult for non-practitioners to understand the underlying logic. But judging from the explanatory framework provided by the author in this book, the key to the development of this technology is to give AI the ability to think. It is precisely based on this development logic that many people in the industry have shown strong interest in the future of AI large models, setting off an upsurge in the application of AI large models.
But at the same time, more and more social scholars and industry insiders are also reflecting on and even questioning the large AI model: First, the data mining and open network applications of the AI large model have a strong scale effect. If AI is unique, If the learning and prediction capabilities are applied to more fields, it may cause intellectual property disputes and involve deeper issues such as data privacy. Second, compared with other solidified computer models, AI large models have strong growth potential and can achieve self-learning and self-evolution under certain conditions. In the long run, they may pose a threat to human intelligence and creativity.
The tool value of large AI models
The birth and rise of large AI models is not only the result of technological progress, but also a great change brought about by the revolution of human knowledge and intelligence. From the perspective of the history of human economic and social development, technological innovation in the industrialization era mainly relies on various types of manufacturing equipment, general machinery and other hardware facilities, with industrialization and urbanization as development paths. In the era of digital economy, industrial digitization and digital industries rely more on software. application, and the large AI model just caters to this trend.
From another perspective, the rapid growth of the global economy in the past was based on productivity improvements and efficiency improvements. However, as the dividends of the third scientific and technological revolution faded, it became more difficult to improve factor productivity, resulting in insufficient innovation vitality and competition. On the contrary, excess restricts global economic growth. In the era of digital economy, the application of large AI models has ushered in unprecedented development opportunities. Although the initial application of large AI models may be more of a tool value, as large model training continues to accumulate and increase, these models and the technical methodologies behind them will likely release greater energy. The author believes that compared to the speed of human education and knowledge update, the iterative optimization speed of large models is much higher than that of humans, so it will expand the boundaries of human knowledge faster.
How to correctly view the tool value of large AI models? This requires returning to the “old problems” of technological and economic and social development. From the perspective of the historical development process, the three industrial revolutions in human history encountered many doubts in their early stages of development. In particular, these new technologies and new applications not only subverted the old production and lifestyle, but also changed the development process of human society, resulting in Social problems such as phased unemployment, unbalanced economic development, and the widening gap between rich and poor have been addressed.
But do these endless problems mean that the fruits of the Industrial Revolution cannot be put to good use? The answer is of course no. As a driving factor, scientific and technological innovation not only has its positive effects, but it also inevitably produces negative problems and even brings serious challenges to economic and social development. In this regard, the author puts forward an interesting point in the book: the contribution of new technologies to human employment often starts at a low level and then goes up. When new technologies first appear, the substitution effect on manpower is more obvious, which will cause resistance from some groups. As the application of technology gradually deepens and spreads, more innovation will be generated, the market size will be enlarged, and then it is possible to create more jobs for mankind.
Therefore, along this line of thinking, to correctly view the tool value of large AI models, two issues need to be clarified. First, the application of large AI models should follow legal system norms. In the process of large-scale application, we should explore and establish corresponding rules, standards and restrictions to ensure that technology applications are controllable, safe and friendly, rather than arbitrary. It grows wildly. Secondly, the development of large AI models should follow its own rules. The ultimate goal of technology application should be to provide more tool support for the development of human society and exist as an auxiliary tool rather than putting the cart before the horse.
New challenges for large AI models
Based on the above analysis, the application prospects of AI large models are very broad and full of imagination. With the emergence of application models such as ChatGPT and "Wen Xin Yi Yan", large AI models have entered the homes of ordinary people from unfamiliar technical fields and triggered various heated discussions. However, in order to truly understand the future development direction of large AI models, We also need to think carefully about the logic and insist on putting people first.
The author believes that enterprises will experience three realms in the application of large AI models: The first realm is "If you want to do your job well, you must first sharpen your tools." Enterprises need to evaluate the application maturity, efficiency improvement, model defects and negative impacts of large AI models; the second level is to "set sail together in the same boat". Enterprises should optimize and adjust large models, paying attention to the impact of large models on employees and conflicts with organizational mechanism design; the third level is "to reach a higher level if you want to see a thousand miles away". Enterprises need to consider business models and existing industry rules, and need to take higher risks.
The "triple realm theory" proposed by the author emphasizes that people's application of large AI models is a step-by-step process, and business logic, social ethics, and legal norms need to be followed in the actual application process. From a realistic development context, any cutting-edge technology may have both positive and negative impacts, and how to grasp the balance is extremely critical. Although society needs to be more tolerant to the birth and development of each new technology, once it develops to a certain scale and has a strong influence, the application specifications and rules of these technologies should be seriously discussed question.
Famous historian Yuval Harari believes that artificial intelligence is mastering language beyond the average human level. By mastering language, artificial intelligence is holding the key to invading human civilization systems. Currently, scholars from various countries and professionals in different fields have also expressed varying degrees of concerns about the evolution of AI. First, AI large model technology is a brand-new and cutting-edge technology. In the Internet environment, these applications are beyond the general understanding of the general public. There is a gap in the existing legal system for the development of AI large models, and non-professionals have even more knowledge. Blind spots; second, the development and application of large AI models has a certain monopoly nature. Currently, there are still very few companies that master these technologies and methods. If they are to be fully introduced to the market and the public in the future, many unknown risks may arise.
Recently, the European Union proposed plans to formulate a new draft law called the Artificial Intelligence Act (AI Act), which includes prohibiting the use of specific artificial intelligence services and formulating relevant legal regulations. At the same time, OpenAI CEO and co-founder also takes a cautious stance on large AI models and said that the application of these new technologies may have risks and challenges.
From the perspective of future prospects, AI large models have huge potential, but there are also unknown risks. How to control this new technology requires full preparation. First of all, we should correctly view the tool value of large AI models, promote the evolution and iteration of large AI models through continuous deep learning, and integrate them into a wider range of application fields. Secondly, we must be fully prepared in terms of laws, systems, and technical specifications, build a more cutting-edge technology supervision system, and promote the orderly, standardized and healthy development of large AI models. Finally, let the large AI model finally return to human value and put people first, rather than replacing people with AI.
Looking to the future, the world is entering an era of accelerated change in large AI models. How to adapt to new uncertainties, how to seize new development opportunities, and how to promote science and technology for good are questions that we must answer when developing large AI models. Looking at every technological change in history, it has not only promoted economic development and civilization progress, but it has also brought about many problems. As the author conveys in this book, large AI models have contradictions between evolution and loss of control, and it is this contradiction that drives us to continue to explore and discover and pursue a better future for human society.