(International Observation) Upstarts in the AI ​​industry "out of the circle": What kind of future does ChatGPT reveal?

　　China News Agency, Beijing, February 12th, title: Upstarts in the AI ​​industry "out of the circle": What kind of future does ChatGPT reveal?

　　Author Zhang Huandi

　　Recently, the chat robot program ChatGPT has exploded and "out of the circle", which has aroused attention and discussion from all walks of life, and the concept of artificial intelligence (AI) has also returned to the public eye.

In response to related topics, Huang Minlie, associate professor of the Department of Computer Science and Technology of Tsinghua University, said in an exclusive interview with China News Agency that the highly versatile ChatGPT has become a milestone in the history of AI development, and how to coexist with AI is a problem that everyone may face .

Why does ChatGPT become an upstart in the AI ​​industry?

　　The United States OpenAI company officially released ChatGPT on November 30, 2022. Only 5 days after its launch, the number of registered users of ChatGPT exceeded 1 million; at the end of January this year, the number of monthly active users of ChatGPT exceeded 100 million, making it the fastest-growing consumption in history. or apply.

　　There are many chatbot programs similar to ChatGPT, but none of them lead the new trend in the AI ​​field like ChatGPT. One of the reasons is that ChatGPT is more versatile and can complete professional tasks such as translation and programming by itself.

"The biggest feature of ChatGPT is the general task assistant, which can complete a lot of open tasks in one model, and also has a very good performance in generating tasks, context understanding, and safety ethics." Huang Minlie pointed out.

　　Technological progress is an important reason for the significant increase in the versatility of ChatGPT.

Huang Minlie believes that compared with the first-generation and second-generation chat robot programs that are mainly based on rule learning, the third-generation chat robot program represented by ChatGPT has made significant technological progress: its large model with Transformer as the basic architecture is The technical base combines a large amount of data and computing power optimization.

　　After ChatGPT, an upstart in the AI ​​industry, became popular, large technology companies such as Google announced the launch of competing products to participate in the new competition in the AI ​​​​field.

Bill Gates, the founder of Microsoft Corporation in the United States, recently said that AI will become "the hottest topic in 2023".

　　"The emergence of ChatGPT is a milestone event in the history of AI development, marking the advent of the era of seamless human-computer interaction. It may reshape the forms of information acquisition such as search, recommendation, and mobile assistants, and have a profound impact on the information industry and the AI ​​​​industry." Huang Minlie said .

　　What controversies does ChatGPT face?

　　The development of AI technology has also brought many controversies.

AI face-changing technology represented by deepfake (Deepfake) has previously triggered a wave of discussions on privacy and truth.

　　In this regard, Huang Minlie believes that ChatGPT has been optimized for legal and moral issues, and its performance is better than that of "peers". When developers collect high-quality human feedback data, they regard issues related to safety and ethics as an important part of ChatGPT The design is in, and through reinforcement learning, ChatGPT has learned the ability to reply safely.

　　Still, ChatGPT faces ethical issues.

Huang Minlie gave an example: If a user asks ChatGPT to take illegal actions, it will be rejected; but if the user asks ChatGPT to provide advice that violates the law under the pretext of creating a novel, they may get an answer.

　　After the launch of ChatGPT, its powerful text writing ability has been favored by students and used to complete thesis.

So far, many universities in the United States, the United Kingdom and other countries have banned students from using ChatGPT on campus.

The American journal Science also recently issued a statement stating that it does not accept submitted papers generated using ChatGPT, nor can ChatGPT appear as a co-author of a paper.

　　According to Huang Minlie, one of the problems with using ChatGPT to complete homework or academic papers is copyright infringement.

But we can’t “give up food because of choking”, “technology can continue to develop, and in the process, we can carry out targeted governance on problems”.

　　On the other hand, scientific and technological progress has failed to solve the problem of unbalanced regional development, and the global technological divide still exists today.

AI is a cutting-edge technology, and only a few countries can invest heavily in this field.

　　Huang Minlie said that large-scale model training closely related to AI requires a lot of computing power and data, and the cost of training and reasoning is also very high. Whether in academia or science and technology, only a small number of institutions can afford the research and development of related technologies. Injustice has been created.

Is the "AI threat theory" alarmist?

　　With the maturity of AI technology, the "AI threat theory" has received more attention.

After the launch of ChatGPT, its various professional functions have been continuously discovered.

There is a view that programmers, translators, customer service and other positions may eventually be replaced by AI.

　　In response to the unemployment problem that may be caused by the development of AI technology, Huang Minlie believes that the popularization of AI tools will lower the threshold for freelance writers, game producers and other occupations, and bring more opportunities for freelance work.

However, after more in-depth integration of artificial intelligence-generated content (AIGC), user-generated content (UGC) and professionally produced content (PGC) in the future, the elimination of some occupations is inevitable. problems to be faced".

　　In addition, the "AI threat theory" in science fiction works often includes an important factor-more and more powerful AI will eventually seek to eliminate or replace human beings.

　　In this regard, Huang Minlie said that the villain AI images in science fiction works have self-awareness, but in reality, AI currently has neither self-awareness nor independent decision-making ability. At this stage, the source of all AI capabilities is the data and feedback provided by humans. .

“Systems such as ChatGPT are typical machine attributes, which exist only to meet people’s information needs or complete tasks, and do not have personality, emotion, style, etc. Machine attributes are related to human-like attributes that can meet emotional, social, and companionship needs. Combination is the real digital life form, that is, 'artificial intelligence' in the true sense." (End)