On July 7, 2023, at the "Global Summit of Artificial Intelligence for Humanity" in Geneva, Switzerland, the simulated robot Sophia (left) talked with the simulated robot Nadine. Photographed by Lian Yi, a reporter from Xinhua News Agency
Elon Musk, a well-known American entrepreneur, said on social media on the 8th that by the end of next year or 2026, the new artificial intelligence (AI) model may surpass human intelligence and be "smarter than the smartest people".
Is Musk's prediction reliable? With the rapid development of artificial intelligence, can human beings enjoy the "dividend"? At the same time, the development of AI will also cause many problems. How to deal with the risks?
Is Musk's prediction reliable?
On the 8th, Musk's interview with Nikola Tangen, CEO of the investment management fund of Bank of Norway, was released on the social media platform X. Musk said in the interview: "I guess that by the end of next year, we will have more intelligent artificial intelligence than any human."
This prospect is limited by related bottlenecks. The Financial Times introduced that the development speed of AI was affected by the bottleneck of microchip supply, especially the microchip produced by Nvidia, which is crucial to training and running AI models. Musk said that although these restrictions are easing, the new model is testing the equipment and power grid performance of many data institutions.
Zhu Rongsheng, a special expert of the Strategic and Security Research Center of Tsinghua University, told the reporter that the power consumed by the human brain is far lower than the power consumed by artificial intelligence during operation, and there is still a gap between artificial intelligence and human intelligence. However, in the long run, compared with the human brain, there are many deficiencies in AI, which will eventually surpass human intelligence under the constant "blessing" of algorithms and data. The hardware and software required to achieve this goal are chips and algorithms respectively.
Last year, Musk established the artificial intelligence startup xAI, which launched its first artificial intelligence model, Grok, in November of the same year. Musk said that the training of the second generation model Grok 2 will be completed in May this year, requiring about 20000 Nvidia graphics processor (GPU) computing chips H100. The training process was once hampered by a shortage of chips. Following Grok 2, xAI will launch a more powerful third-generation model, which is expected to require more than 100000 chips. He said that the main constraint to the development of artificial intelligence last year was the lack of high-performance chips. This year, the supply constraints are turning into voltage transformers; In a year or two, the limit will be power supply.
International cooperation is needed to collect "dividends"
In the past 18 months, breakthroughs in artificial intelligence, including video generation tools and chat robots, have greatly accelerated the development of artificial intelligence.
The Financial Times reported that Musk predicted last year that mankind would "fully" realize general artificial intelligence by 2029. He has always been optimistic about the development of the so-called general artificial intelligence, believing that this kind of artificial intelligence tool is so powerful that it can defeat the most capable human individual in any field. Demis Hassabis, co-founder of Deep Thinking, Google's AI company, predicted earlier this year that general AI could be realized by 2030.
Zhu Rongsheng said that the development of AI first needs policy, resources and technical support, and the "cultivation" of AI in many fields lacks relevant support; Secondly, cooperation between countries is needed, but at present many international cooperation is constrained by geopolitical and other factors, and the development of artificial intelligence is constrained; Third, the development of AI is constrained by cultural characteristics. For example, the answer given by ChatGPT is more in line with the English context. In this way, developing AI with a single national force may increase the imbalance of global AI development, that is, the data gap. To solve the above problems, it is necessary to link the funds with the market accurately and carry out extensive international cooperation.
Eric Salheim, former Deputy Secretary General of the United Nations and former Executive Director of the United Nations Environment Programme, said in an interview with Xinhua News Agency recently that the world needs to carry out global cooperation on artificial intelligence, which requires close cooperation between China, the United States, European countries and other major countries to develop reasonable laws and regulations, so as to seize the opportunities brought by new technologies and benefit from them.
How to deal with risks?
AI has reduced production costs, improved production efficiency and opened the door to new formats in many industries. Although the efficiency has been improved, the use of AI has caused many problems and risks.
According to a survey of scientists conducted by Nature in 2023, 30% of respondents admitted to using AI tools to help write articles. According to the website of Popular Science, many researchers rely on AI tools, many professional journals are full of "nonsense" generated by AI tools, and many articles use AI obviously.
As early as 2014, more than 120 articles were deleted from the journals of the American Institute of Electrical and Electronics Engineers and the German Springer Press, because these articles contain "nonsense" generated by artificial intelligence. Popular Science commented that in the next 10 years, as more complex and easier to use tools such as ChatGPT of OpenAI have been widely used, the number of AI generated texts in various journals will almost certainly increase.
Zhu Rongsheng believes that the social problems caused by artificial intelligence spread faster and more widely, and this technology continues to increase the complexity of legal issues involving copyright law, copyright, privacy and other legal issues, so the moral boundaries and legal norms applicable to new technologies are looming.
"The era of artificial intelligence is an era where risks and opportunities coexist. This technology may bring more uncertainty than certainty to human society and destiny," Zhu Rongsheng said. "In the face of this irresistible wave, we need to maintain a calm and prudent attitude and thinking in the AI boom."
(Reporter Ge Chen, Wang Yijun, Yin Xiaosheng)
Xinhua News Agency, Beijing, April 10
