October 13, according to the Korean media, The Korea Herald, an artificial intelligence pioneer, founder and chairman of AMI LabsYang Likun（Yann LeCunOn Monday in Seoul, he wrote:AI THE NEXT BREAKTHROUGH WILL COME NOT FROM LARGE LANGUAGE MODELS, BUT FROM SYSTEMS THAT UNDERSTAND AND PREDICT THE PHYSICAL WORLD. HE BELIEVES THAT THE WORLD MODEL WILL DRIVE AI INTO THE NEXT PHASE。
Yang Liqun stressed that there are limitations in understanding the real world in the larger language model developed for everyday life. “We don't have robots who can do their job without any training,” he explainsI DIDN'T DEVELOP AN L5 SELF-PILOTIn contrast, any young person can learn to drive only a few hours of practice."
AND HE WENT ON TO SAY, "THIS MEANS THAT THERE'S A HUGE GAP BETWEEN UNDERSTANDING LANGUAGE AND UNDERSTANDING THE REAL WORLD. IT HAS PROVED THAT IT IS EASY FOR MACHINES TO READ LANGUAGES AND NOT EASY TO UNDERSTAND THE REAL WORLD.”
THE BIG LANGUAGE MODEL (LLM) IS A MODEL BASED ON BIG TRAINING DATA,Predicting the most probable next word generation textI don't know. In the view of Yang Liquon, this approach is performing well in tasks involving predictable sequences, such as programming, mathematics and so forth, but most of the information in the real world is far less predictable than in language。
That's why he advocated a world model,TO ENABLE AI TO LEARN THE PHYSICAL PATTERNS AND CAUSALITY THAT DOMINATE THE REAL WORLDTo allow the system to predict the likely results of the operation and interact with the surrounding environment。
NOTE: YANG LIQUON, ONE OF THE MAJOR INVENTORS OF THE CONGESTED NEURAL NETWORK (CNN), HAS A THEORY THAT GREATLY PROMOTES MODERN COMPUTER VISUAL DEVELOPMENT AND IS WIDELY USED IN IMAGE AND VISUAL RECOGNITION. HE WAS AWARDED THE TURING PRIZE IN 2018 AND BECAME ONE OF THE “AI GODFATHERS”。