On December 14, the 2023 "Nature's 10" were officially announced. In addition to the 10 people selected from major global scientific events - 5 men and 5 women, there is also 1 non-human on the list this year - ChatGPT, a chatbot released by the American artificial intelligence company OpenAI.
In fact, ChatGPT should be among the top 11 figures in Nature. This in itself has changed history and also changed the rules of evaluation and selection of Nature. Changing the rules of "Nature" is also reflected in the fact that ChatGPT is not a natural person, but "Nature" treats it as a natural person.
Of course, ChatGPT is not just changing the rules of "Nature" in history, but like the other top 10 figures, it may have a huge impact on human society and change the course of history. The top 10 people selected by "Nature" are actually no less than the Nobel Prize winners every year, because they are all outstanding and outstanding contributors in various fields, and their scope is wider than the Nobel Prize, including nature, humanities and social fields. For example, one of the top 10 people this time is Brazil's Minister of Environment Marina Silva (Marina Silva), on the grounds that she protects the Amazon rainforest.
As for why ChatGPT was selected as one of Nature's top 11 figures, the magazine's editorial board reasoned: "Although (ChatGPT) is not a tool and does not fully meet the selection criteria for Nature's top 10 figures, we made an exception and included it in the list to recognize the huge changes that generative artificial intelligence has brought to scientific development and progress."
In fact, ChatGPT is more famous and influential than any figure selected by Nature, and will have a greater and far-reaching impact on future human society. Therefore, ChatGPT should be regarded as the first of the 11 big figures. Since ChatGPT was released as a free-to-use automated conversational agent (chatbot) in November 2022, it has been occupying various news headlines around the world. Not only the scientific community, but the entire society has personally felt its shocking impact.
ChatGPT is not one person, but it is tens of millions of people, the gathering and projection of countless people. The American company OpenAI created ChatGPT. Its foundation and building materials are neural networks with hundreds of billions of parameters. This neural network is built on and trained on a huge corpus of online books and documents, and is estimated to cost tens of millions of dollars. There are also a large number of researchers and editors who evaluate its responses to further refine and shape the content of its output. In 2023, OpenAI upgraded ChatGPT’s underlying large language model (LLM) and connected it with other programs so that the tool can receive and create images.
This also means that ChatGPT is by far the most powerful and knowledgeable encyclopedic "giant". It is better than countless professors, scholars, engineers, writers, doctors, pharmacists, accountants, actuaries, designers, journalists and almost all human professionals. Therefore, there is a good reason for "Nature" to list it as the eleventh person of the year. Who can beat or be more powerful than ChatGPT, a neural network with hundreds of billions of parameters?
Although ChatGPT looks very powerful, from a certain perspective, it is just a giant with clay feet, and it is not as good as a living natural person in dealing with many simple problems. There are many reasons, such as lack of background, limited knowledge to train it, etc., but the most important thing is that the knowledge and training it acquires and the information it outputs are uncertain, because it does not have the ability to discern human society and nature like humans, as well as the ability to reason logically, and of course it lacks human emotions. It will not be like a primary school teacher who will say with absolute certainty that 3+2 equals 5 when teaching students. If you tease ChatGPT and say, my wife said 3+2 equals 6, my wife must be right. Therefore, ChatGPT will immediately change its mind and say, sorry, I was wrong, 3+2 should equal 6.
Of course, the important role and great functions of ChatGPT have been recognized by humans, and it will become one of the most important tools for humans. Whether it is helping people write and produce various products, such as various texts, writing programs and codes, creating pictures, designing drugs and vaccines, etc., it is both fast and of a certain level. However, the limitations and shortcomings of ChatGPT are also obvious.
Interestingly, one of the 11 big names in "Nature" this time also includes Ilya Sutskever, the creator of ChatGPT. Sutskevi is wary of ChatGPT and other AI products, so he is relatively conservative in treating the ChatGPT he created. He believes that ChatGPT and other AI may cause huge risks and disasters if they are not effectively controlled by humans. As a result, he is now focusing on controlling artificial intelligence systems through OpenAI's "Hyper-Alignment" project. Moreover, he believes that OpenAI and some other companies should keep their code and training data secret. In the long run, closed systems will prevent others from creating powerful artificial intelligence, which is the responsible approach. "At some point, the capabilities of AI will become so vast that open source models will be patently irresponsible."
Sutskowe's views were clearly at odds with OpenAI CEO Sam Altman's, which sparked a feud within OpenAI, with the latter first being fired and later reinstated. Ultraman strongly advocates the rapid introduction of ChatGPT to the market. However, this time, "Nature" ranked Sutskvi instead of Ultraman as the 11 major figures, which reflects the thinking and trend of "Nature". ChatGPT and other AI products can be developed and used, but they must be cautious.
Because, in addition to serious nonsense, ChatGPT also has other errors and hidden worries, including helping cheating and plagiarism, helping fraud, fabricating facts, inventing references, copying hate speech, spreading rumors, etc. In addition, ChatGPT will play any role desired by its interlocutor, sometimes submissive, sometimes charming, sometimes funny, sometimes scary.
All of these are problems that arise when humans design, train, and use ChatGPT. If the products and answers provided by ChatGPT are not manually reviewed and verified, errors may occur, and the errors may be outrageous. Therefore, for ChatGPT and other AI software and tools, in addition to constantly improving and improving the technology to suit current human cognition and actual conditions, legal supervision is also required.
On December 8, 2023, the European Commission, the European Parliament and representatives of EU member states reached a preliminary agreement on the Artificial Intelligence Act (AIAct). The new bill involves general-purpose AI systems including ChatGPT. This is the world's first comprehensive regulation for AI supervision, aiming to find a balance between protecting basic human rights and not hindering the development of the artificial intelligence industry. The bill classifies AI systems into different levels such as low risk, high risk, unacceptable risk, and specific transparency risk, and formulates corresponding regulations according to risk levels.
Therefore, "Nature" lists ChatGPT as the 11 big names of this year not only because it will become mankind's best helper, but its application is also one of the biggest and most difficult challenges for mankind now and in the future.