Technology gurus debate big model: How far is China from world-class and how to seize the opportunity of a century

The big model that has been "raging" all the way has become the focus topic of this year's World Intelligence Conference.

On May 18, during the 7th World Intelligence Conference, Robin Lee, Zhou Hongyi, Liu Qingfeng, Wang Jian and other Internet entrepreneur representatives unanimously chose the big model as the theme speech. Several domestic technology giants, including Baidu, Tencent, Alibaba, Huawei, JD.com, and 360, have also announced corresponding layouts in the generative AI industry, and the battle over big models is becoming increasingly fierce.

At the conference, Wang Zhigang, Minister of Science and Technology, stated that he will closely seize the new opportunities of global artificial intelligence development, respect the laws of technological innovation and artificial intelligence development, increase the layout of research and development of basic theories and cutting-edge technologies in artificial intelligence, create a number of regional highlands and technology platforms for artificial intelligence, deepen the integration of industry, academia, research and application led by enterprises, and promote the empowerment of economic and social development by artificial intelligence.

Wang Zhigang, Minister of Science and Technology, Visual China Data Map

The future growth potential of the large model is also worth looking forward to. According to the calculations of the China Academy of Information and Communications Technology, the scale of China's artificial intelligence core industry reached 508 billion yuan in 2022, a year-on-year increase of 18%, and the number of enterprises exceeded 4000. China's artificial intelligence industry has entered the world's first tier.

People have achieved from discovery to creation through computationally driven technological innovation. "Alibaba Cloud founder Wang Jian reviewed several key points in the history of technology development on site: DeepBlue in 1997, AlphaGo in 2016, and ChatGPT this year. In his view, there are still infinite possibilities for the future of big models.

Benchmarking ChatGPT: How far is it from world-class?

In the era of big models, there is a special phenomenon: globally, only two countries and regions have the core capabilities of this industry, North America and China, which can truly make big models. "When discussing the current development status of domestic big models, former Baidu Vice President, founder and CEO of Qiji Innovation, Lu Qizeng, said.

Lu Qi Visual China Data Map

ChatGPT, a chat robot developed by artificial intelligence company OpenAI, has become popular worldwide, triggering a new round of artificial intelligence "arms race" and becoming a benchmark for domestic internet companies.

At present, iFlytek Spark is only one step away from ChatGPT, only two points away from 100 points, and will surpass it before October this year. In a live speech at the World Intelligence Conference, iFlytek Chairman Liu Qingfeng talked about the distance between its iFlytek Spark model and ChatGPT.

Previously, on May 6th, iFLYTEK officially released the iFLYTEK Spark Cognitive Model. Liu Qingfeng stated that the current iFLYTEK Spark Cognitive Model has surpassed ChatGPT in three major abilities: text generation, knowledge Q&A, and mathematical ability. He mentioned that on October 24th, a universal model benchmarking ChatGPT will be implemented, with Chinese surpassing the current version of ChatGPT, English being able to achieve a considerable level, and leading the industry in fields such as education and healthcare.

Despite constant rhetoric, the consensus in the industry is that there is still a gap between domestic models and international first-class models, and surpassing them is not an overnight achievement. 360 founder Zhou Hongyi said, "I have no doubt about the ability to create large models, but we also need to be aware that there is still a significant gap between domestic large model products and GPT-4." In his view, if we don't go through these two years of imitation and plagiarism and say we can surpass them, it's called "bragging".

He mentioned that domestic athletes initially adopted a strategy of imitation and following, but in the process of gradually investing in research and development, as the scene expands and users expand, they will become more aware and may even achieve overtaking on curves and catching up in the future.

Tencent CEO Ma Huateng also holds a similar view. At the previous Tencent shareholders' meeting, Ma Huateng believed that truly building a big model requires more accumulation and sedimentation, and should not be rushed. He said that artificial intelligence is a rare opportunity in the internet industry, similar to the industrial revolution that invented electricity. Internet companies have a lot of accumulation in the AI field, and Tencent is also immersed in research and development, but is not in a hurry to complete it early and showcase semi-finished products.

Ma Huateng Visual China Information Map

For the Industrial Revolution, taking out the light bulb one month early is not that important in the long term. The key is to solidly do a good job in the underlying algorithms, computing power, and data, "Ma Huateng said.

In Wang Jian's view, what truly drives the development of big models is creativity. The true computing power is not simply asking for 100 well structured A100 cards, but rather whether we truly have the ability to write creativity and ideas into the rules once again in human history. This is what we truly need to surpass in the next 5-10 years

How to ensure AI security? To explore in development

The development of artificial intelligence requires autonomous and secure computing power. The next step in the development of artificial intelligence will involve ideas and ideas, which is a problem of autonomy and controllability for any country. "At the World Intelligence Conference, Zeng Yi, the Chairman of China Electronics Information Industry Group Co., Ltd., believed.

How to control the potential risks while AI technology is rapidly advancing? Multiple governments have introduced corresponding regulatory measures. Sam Altman, CEO of OpenAI, also called on the government to establish security standards for AI systems and stated that 'government regulatory intervention is crucial'.

Zeng Yi stated that computing power is the most urgent and significant issue to be solved in the next stage of artificial intelligence, digital China, and digital economy. China Electronics has made certain progress in independently controllable, soft hard collaborative, and vertically penetrating computing and security integrated computing power base. In terms of data elements, China Electronics has conducted some theoretical research and established China Electronics' data group.

In his view, the applications surrounding autonomous and controllable scenarios have not yet fully opened up. Through digital government platforms, digital ministerial platforms, digital central enterprise platforms, and other digital private enterprise platforms, it is necessary to develop application scenarios as soon as possible.

Zhou Hongyi admitted in a written interview with a reporter from Pengpai News that there will definitely be many insecure issues with large models, mainly reflected in two aspects. One is their own security issues, such as software vulnerabilities, algorithm/model vulnerabilities, data security, etc; The second is the security issues that arise, such as being controlled, maliciously applied, abused, and misused, leading to security issues.

It's a bit metaphysical to talk about big model security issues now, asking questions but not finding a solution. Because we don't know much about ChatGPT, and our definition of ChatGPT is also based on our early understanding. Today's ChatGPT is a true industrial revolution level technological change. The industrial revolution means that it will bring significant changes to the entire country, the entire world, and humanity, Not developing is the greatest insecurity. Security issues should be considered during the development process and solutions should be explored.

Zhou Hongyi Visual China Data Map

Earlier, Tencent President Martin Lau also mentioned the regulatory issue of generative AI, "We believe that regulation is necessary. This is not only in China, but also a global problem. For example, in the United States, there is still a lot of discussion about regulatory issues, and even the founder of OpenAI participated in the hearing to require regulation of the industry. Under the appropriate regulatory framework, the industry will have healthy growth space."

Will AI take away human jobs?

The rapid development of AI has also raised concerns about its replacement for human professions. According to a report released by investment bank Goldman Sachs, with the breakthrough of AI technology, it is expected that 300 million jobs worldwide will be replaced by generative AI.

It is worth noting that at the World Intelligence Conference, technology leaders generally expressed an optimistic attitude towards this issue, which alleviated the "spiritual internal friction" of migrant workers.

Robin Lee, CEO of Baidu, said that AI will greatly improve people's work efficiency. The reason why people are worried that their careers will be replaced by AI is that they can see that the current jobs will disappear, but they cannot see that new jobs will be created.

"Just like those people 100 years ago and 200 years ago, they could not see the new job opportunities created later. Personally, I am an optimist, and I am not worried that the big model will lead to fewer job opportunities and worse life for human beings." Robin Lee said frankly.

Previously, Zhou Hongyi also talked about the potential unemployment issues brought about by artificial intelligence, which he believed was "creating anxiety". In his view, human decision-making, emotions, imagination, and creativity are irreplaceable in the short term by artificial intelligence. The artificial intelligence big model is the greatest tool invented by humans, which embodies the knowledge achievements of all humanity and can endow ordinary people with stronger abilities. In the future, those who cannot ask questions or use artificial intelligence big models are more likely to be eliminated.

"Historically, the development of the industrial revolution has led to the disappearance of some jobs, but it has brought more job opportunities. A hundred years ago, the coachman disappeared, but there were jobs such as drivers and power maintenance workers; 30 years ago, typists disappeared, and instead were network analysts, programmers and game developers." At the scene, Robin Lee expressed his attitude by tracing back to history: AI will not reduce job opportunities, But it will bring the next growth miracle to the global economy.