The capabilities of China’s own large language model are two to three years behind those of GPT-4, a Chinese tech mogul said during the China Development Forum 2023 on March 27.
Zhou Hongyi, founder of cybersecurity software powerhouse 360 Group, told the audience that when OpenAI’s immensely popular conversational AI tool is further updated to GPT-6 or GPT-8, a new “species” will be born, with its own cognitive ability and true sense of self.
His claims contrasted sharply with Robin Li, founder and CEO of search engine giant Baidu, who announced recently that Baidu’s ChatGPT-like application Ernie Bot is only “one to two months” away from catching up with GPT-4.
In a more pragmatic tone, Zhou said there exists no insurmountable technological barriers to develop China’s answer to ChatGPT now that the way forward is clear.
The country thus should adopt long-termism and try to come from behind in the field of LLM-based AI innovation, he noted.
Zhou is among a number of notable business leaders who attended the forum, alongside Tim Cook of Apple and Ray Dalio of Bridgewater Associates, among others.
Zhou began his speech by presenting the audience with an AI-generated self-portrait and introductory remarks.
In his speech, he divided AI’s development into three phases characterized by three “s,” namely, stupid AI, strong AI and super AI.
In his view, right now we are at the stage of a true AI revolution in the sense that disruptive innovations, represented by GPT-4, possess attributes that are well beyond mankind.
When OpenAI upgraded ChatGPT to GPT-3.5 in January, a fine-tuned version of GPT-3, the Chinese internet magnate predicted that the chatbot program will soon have its own “eyes and ears,” meaning ChatGPT can recognize pictures.
The scale of the technological progress within the past several months has proved him right, with GPT-4 being able to process and analyze images.
Zhou’s words came at a time of rising concerns about humans being replaced at a faster rate by robotics and AI, in not just tedious, repetitive tasks, but also those in white-collar jobs.
At the forum, he openly acknowledged the risks faced by mankind, but emphasized that in the future, people at risk of losing their jobs to AI are those who are unable to ask relevant questions or do not know how to use LLM.
“LLM is by far the greatest tool humans have ever invented,” said Zhou.
According to him, LLM may evolve and adapt on its own, posing unpredictable security challenges to man.
On the upside, as AI continues to develop, humans likely will realize the once impossible scenario of digital immortality, where we will be able to converse with ancient sages and preserve the human civilization in a better way.
