BEIJING, December 4 (TMTPOST) – Chinese computer scientist Zhang Hongjiang delivered a keynote speech on AI-generated content (AIGC) during the 2023 T-EDGE Conference last Friday, exactly one year since the official launch of ChatGPT.

Zhang is an international academician of the National Academy of Engineering in the United States and the director of the Academic Advisory Committee of the Beijing Academy of Artificial Intelligence (BAAI).

Under Zhang's leadership, the BAAI has become one of the most attention-grabbing Chinese AI large model research institution in the international AI arena. It introduced the Wu Dao large model, pioneering large model research in China.

Zhang delivered a speech titled "AI Large models Drive a New Paradigm in Industries" from California to the 2023 T-EDGE Conference last Friday, elaborating on the potential impact of AIGC on various major industries.

During the speech, Zhang reviewed the large model craze brought about by GPT and shared his observations and thoughts behind it. He also examined how large models drive a new paradigm in industries and change the entire artificial intelligence (AI) industry chain. Finally, Zhang shared his outlook on large model technology, especially how large models will change technology development trends and people's lives.

Zhang said that ChatGPT is the "watershed" of the third wave of AI for humans. With the iteration of technologies like GPT-3.5 and GPT-4, an AI system capable of precisely understanding language logic finally came.

"Today's OpenAI is not just an AI technology company but also an AI platform company," Zhang said in the speech.

Besides, he said that AI large models, as the foundational platform, will systematically drive all industries into a new paradigm, rewriting all software applications and industries, and enhancing productivity in every field. Additionally, utilizing technologies like GPT Copilot as future organizational forms will significantly improve company efficiency, leading to fundamental changes in organizational structures.

"Large models will reshape all industries and enhance productivity in all fields. With continuous improvement and the autonomy of existing models, they are expected to change the entire industry landscape," said Zhang. In the next 1.5 to 2 years, humanity may witness the dawn of large-scale commercialization.

Zhang also admitted that the "illusion" of large models is currently a big problem. The large model illusion problem refers to the generation of inaccurate, incomplete, or misleading outputs by some artificial intelligence models when faced with certain inputs. Although the latest GPT-4 has made significant improvements in addressing the "illusion," with a 40% improvement compared to ChatGPT, the "illusion" rate is as bad as around 10%-14%. Therefore, enterprises and developers need to continuously improve the capabilities of large model technology to truly implement it in industrial scenarios.

"What I want to emphasize is that although we have a large model that is doing well in some application scenarios, we should not ignore the need for continuous investment, improvement, and development in general large model capabilities. Otherwise, perhaps in another 6 months or 12 months, many applications relying on large models will be abandoned by users because they cannot meet user expectations," Zhang said. All software companies must embrace large models, including those focusing on software tools and application services. They all need to employ large models to rewrite software.

Looking to the future, Zhang quoted a study from OpenAI, which shows that 80% of American jobs will be affected by AI technology, especially creative jobs such as those related to audio and image generation. This means that highly paid "white-collar" workers will be more significantly affected. Meanwhile, AI has surpassed 85% or 90% of human candidates in various types of exams. This will affect the development of professional fields such as law and medicine.

"Large models will definitely have an impact on human work and the future of all humanity," Zhang said. AI large models will not only replace "white-collar" jobs but also drive improvements in fields such as robotics and autonomous driving.

"In the future, we will usher in a world of 'autonomous intelligence,'" Zhang concluded in the speech.