Vice-Minister for Foreign Affairs

As the first doctoral student of Giorgi Parisi (Giorgio Parisi), it is difficult to keep up with the pupil and has not yet been understood by the instructor, since the subject of the Pacife study is extremely cross-cutting, although he has had eight years since 1981 to learn and work with Pairis.

The 74-year-old Palis received the 2021 Nobel Prize in Physics for discovering the interaction of physical system disorders and fluctuations from atom to a planetary scale. His research was sufficiently complex: from basic particle physics to machine learning, from cycling glass to water hanging, from stock markets to weather, from design experiments to more efficient cooking Italian.

The current director of the Swiss Centre for Data and Network Science, who was awarded an award of stipulation in recognition of the complexity of the scientific research and the maturity of Paris, meant that the complex science was placed in the room and was no longer an marginalized role.

In recent days, the Chinese Science Journal conducted a special visit to Parisi to comment on the views of the ChatGPT and the future development of artificial intelligent capacities in the immediate aftermath of the anomaly.

The Chinese Science Journal: The winner of the No Award, Philip Andersen, has said a word “multipleally different”, that is, the increase in the number of components of a system that not only determines systemic changes but also determines their qualitative changes. Does this sentence also apply to the machine learning of ChatGPT or even human learning?

Parissa: Over the past 35 years, I have studied machine learning without examining recent developments in detail, but I have read scientific and technical papers on what ChatGPT is doing.

The large-scale language model (LLM), which has been described, is able to learn many texts, like human beings, but there is no inherent method of checking the reliability of what it does. I mean that everything that has taken place is simply “unmanufactured”, as a high school student does not know the answer when the teacher asks questions, but “invents” the right answer. For example, I asked friends: what should be read if an Italian poet’s paper is to be written? He gave me a paper, but none of the papers were available. The problem of large-scale language models is similar, in which there is no understanding of reality only if the language is understood.

ChatGPT, of course, is useful for many tasks. If you write a very long paper and wish to summarize it, you can ask ChatGPT. While some misperceptions may arise, I think it will be very good. ChatGPT can therefore effectively avoid duplication of tasks, but it is far away from something similar to that of intelligence.

China Science: Can you describe the future of manual wisdom? Will this not be the same as the slide films?

Parissa: I think it is difficult to predict the future. Understanding is already difficult and predicting the future is more difficult.

The manual intelligent system will free us from duplicative work. For example, automatic translation is available if you simply want to read articles and newspapers or to understand scientific and technical papers. However, if you want to understand in depth an original paper, it would be preferable to read the original paper itself rather than the ChatGPT summary. Like automobiles, manual intelligence has a long way to go.

China Science: you have repeatedly referred to the “right” and “tacit” of the human being in the book accompanying birds. So can artificial intelligence develop a sense?

Parissa: I do not see the way in which manual wisdom is developed. If you want to have “direction”, there must be institutionalized images of the world that are non-formal, and that need to be used, on a case-by-case basis, to use and form networks.

In many cases, you need a analogy that compares one or the other, all of which are related to an inherent awareness of the world. At present, manual intelligentness has not advanced in this direction. Moreover, the number of sudden hits in the current real nerve network is much smaller than in human minds, and a much larger number of sudden hits are required if manual wisdom is to be developed. While this takes considerable time, I am confident that it will move in the right direction.

China Science: Holkin has said that if the robotic continues to develop, it can destroy humanity. How do you look at his concerns?

Parrysi: I do not believe that manual wisdom will destroy humanity, and I think that human beings are more likely to destroy themselves through a nuclear war. Another extremely dangerous issue is climate change and pandemics and limited energy resources, which we may pay a heavy price for competing key resources, which can also lead to war.

That is why I think that the most dangerous thing facing humanity is self-毁灭 rather than the destruction of artificial intelligence. At present, everything in hand is within our control, but it is difficult for the future to say that it is impossible for me to think that artificial intelligent destruction of human beings is impossible.