Looking forward to digital intelligence, the ＂Hundred Models War＂ tends to a new stage of rational generative artificial intelligence.
Inventory Prospect Generative Artificial Intelligence
editorial comment/note
Recently, the National Language Resources Monitoring and Research Center released the "Top Ten New Words" in 2023, among which "generative artificial intelligence" ranked first and was highly eye-catching. As a hot term throughout the year, generative artificial intelligence has benefited from the emergence of ChatGPT since the beginning of 2023. This has also led to the rapid development of AI algorithms, big models and big data, and many large technology companies at home and abroad have come off the AIGC track. After a whole year’s competition, AIGC has now presented the market grand occasion of various language model products and applications, and was once considered as a key variable and an important symbol to promote artificial intelligence from 1.0 to 2.0. Its development prospect in the new year is worth looking forward to.
Forward-looking of mathematical intelligence
■ China Economic Times reporter Hu Pan
2023 is destined to be a year of surging artificial intelligence technology. This year, the interactive AI application "Let a Hundred Flowers Bloom" has become a topic of repeated discussion and reporting by the public and the media. Among the "Top Ten New Words" in 2023 released by the National Language Resources Monitoring and Research Center recently, "AIGC" impressively ranks first in the list, which reveals that AIGC has brought an impact that cannot be ignored in China throughout the year, and on the other hand, it also shows that public opinion in the whole society is expecting the development trend of AIGC in the new year.
"A hundred flowers blossom", AIGC enters a new stage.
Throughout 2023, from the "Hundred Modes Battle" competition to the "Hundred Flowers Bloom" of the application layer ecology, the AIGC industry ushered in an outbreak in this year.
In the international field, the AI big model shows its edge and is controlling the future with technology. In this year, OpenAI officially announced the multimodal model GPT-4, adding image function (multimodal), which changed it from a conversational Demo to a super APP that can multi-line process text, images, voice, video and other information carriers. At the same time, a number of technology giants such as Google and Microsoft have launched their own AI models in order to seize market share.
At home, the "Hundred Models War" is also being staged. Baidu’s ERNIE Bot, Alibaba’s Tongyi Qianwen, 360′ s 360′ s Zhinao, Tencent’s Hunyuan, Huawei’s Pangu … This year, the domestic big model also showed a scene of hundreds of battles. According to the data, as of October, China has released 238 large models, which has tripled compared with 79 in June.
It is worth mentioning that in August this year, the Interim Measures for the Management of Generative Artificial Intelligence Services issued by seven departments, including the National Network Information Office, was officially implemented, and the entry threshold was set from the policy level. The standard system covers many industries, including model industry capabilities, model engineering performance, model computing network, and model safety and reliability. The importance of data, algorithms, models and safety factors is prominent. This also means that the "Hundred Models War" has returned to rationality, and the AIGC industry has entered the stage of integration.
"In this process, how to learn from the principles of biology and psychology, so that the AI system can adapt to new tasks and environments more quickly through meta-learning; How to reduce the dependence on a large number of labeled data, so that the AI system can learn knowledge through unsupervised or self-supervised; How to make AI system process and integrate information from different modes (such as images, voices, texts, etc.) and improve its understanding and application ability; How to reduce the scale and computing resource requirements of AI model in order to deploy and run on edge devices; How to prevent and respond to attacks against AI systems, such as countering samples, model theft and data poisoning, and ensuring the security and reliability of AI systems are all issues that need to be discussed urgently in the development of AIGC technology. " Shenyang, a professor at Tsinghua University School of Journalism and director of the Meta-cosmic Culture Laboratory, said bluntly when interviewed by a reporter from China Economic Times.
"It is also huge and small", and the risk points of AI development still need to be paid attention to
With the rapid updating and iteration of the AI big model, what trends will the AIGC industry show in the future and what sparks will it collide with?
In this regard, Shenyang said that the future development of large language models will present two opposite trends, namely, huge model and micro model. Among them, the giant model is full-modal, with massive parameters, that is, it draws wisdom from a wider range of knowledge, puts forward deeper opinions on problems, emphasizes the infinity and diversity of knowledge and the role of AI technology in pursuing truth and understanding the world. Micro mode is model compression and optimization, and strives to achieve approximate performance in limited resources. That is to say, while maintaining high prediction accuracy, the model size and calculation cost are greatly reduced, the calculation efficiency is optimized, and efficient model prediction is realized, which can be run on smaller equipment.
"In view of the global language and cultural differences, data security and compliance requirements, the localization of large-scale language models is just needed. Localization can improve the accuracy and practicability of the model in a specific language, culture and industry background, thus meeting the needs of more users. At the same time, there will not be only one big model in China. Every enterprise and individual will have their own GPT, which will present a pattern of multiple symbiosis. " Shenyang said.
It should also be noted that there are still some risk points worthy of our attention in the development of AIGC.
Shenyang said that AIGC is strongly involved in the global industrial chain, or it will completely replace programmers, graphic design and legal AI customer service, and set an upper limit for labor costs. The demographic dividend of third world countries no longer exists. The industrial chain of the third world will suffer a huge impact, and the low-end industrial chain will no longer be transferred. AIGC supported by large computing power may become a sharp edge to split the global industrial chain of multinational companies.
At the same time, under the influence of technology Matthew effect, useless classes will be formed predictably in a stable society. Most people’s mediocre mental work will lose its value.
In addition, the ethical risk of AI will be multiplied with the strength of AI technology, thus affecting more people.
In view of this, Shenyang suggested that the future development of AI should also introduce ethical and humanistic care, develop personalized AI that can reflect users’ diverse values and needs, design AI education that can improve people’s information literacy and critical thinking ability, and establish laws and regulations that can prevent AI abuse and protect users’ rights and interests, so as to avoid AI risks from multiple dimensions.
The copyright of this WeChat official account belongs to China Economic Times. If you reprint or quote the contents of this article, you must obtain permission, and indicate that it was transferred from China Economic Times.
Reporting/feedback