A public debate between two leading figures in the field of artificial intelligence, Demis Hassabis and Yann LeCun, is heating up the major question of the technology industry whether general artificial intelligence (AGI) is really feasible or not?
The debate not only reflects academic differences, but also shows opposing directions in how technology giants visualize the future of AI.
AGI has become a central topic since ChatGPT created a global wave in 2022. Recently, Yann LeCun, Professor at New York University, former Director of Meta AI and Turing Award winner, attracted attention when he appeared on a podcast and declared that the concept of AGI is "wrong".
According to LeCun, human intelligence is not general but specialized, closely linked to the ability to physically interact with the real world.
LeCun believes that humans may seem versatile, but in reality they are still seriously limited. We cannot understand and solve all problems, and are not even aware of all the challenges we are facing.
Therefore, he argues that expecting a general AI system to be able to do everything is just an illusion.
Refuting this view, Google CEO DeepMind Demis Hassabis said that LeCun has confused general intelligence and total intelligence.
According to Hassabis, generalization does not mean that a system can do everything perfectly, but rather the ability to learn and solve many different types of problems in theoretical principles.
Hassabis cited the concept of Turing Machine to explain that, in theory, a general intelligence system can learn anything that can be calculated, if there is enough time, memory and data.
He believes that the human brain as well as current AI platform models are "not perfect in reality" due to time and memory limitations. Therefore, AGI is not impossible.
This exchange quickly attracted the attention of Elon Musk, founder of Tesla, SpaceX and xAI. Mr. Musk shared Hassabis' point of view and publicly expressed his agreement, saying that DeepMind leaders made the right argument in the debate with LeCun.
According to the common understanding in the industry, AGI is a form of artificial intelligence capable of learning, adapting and solving completely new problems, going beyond the scope of training data.
An ideal AGI system can improvise in real time and handle unprecedented situations, similar to or even beyond human intelligence.
However, LeCun emphasized that current AI systems, including ChatGPT or Gemini, are still very far from that goal.
Although they can overcome difficult exams and handle many complex tasks, they still lack the ability to deeply understand, be flexible and connected to the real world, which are the core elements of human intelligence.
The debate between Hassabis and LeCun, along with the appearance of Elon Musk, shows that AGI is not only a technological issue, but also a philosophical and fundamental science question, with no final answer.
