According to news on October 20, Yann LeCun, chief artificial intelligence researcher of social media Facebook parent company Meta, said in an interview that artificial intelligence is now stupid and does not need supervision at all. He claimed that early attempts to hinder artificial intelligence would stifle competition.
Yang Likun is the winner of the 2018 Turing Award and a leader in the fields of computer vision and neural networks. His work is often cited by other artificial intelligence researchers to explain why the field of artificial intelligence is currently experiencing great success. Yang Likun compared the current regulation of artificial intelligence to the early days of government interference in the Internet. He said limiting artificial intelligence now would be like regulating the jet before it was invented.
Yang Likun believes that the current supervision is centered around the "potential risks" of the system, but in fact "we currently don't even have a system with a learning ability comparable to that of a cat." He called the approach "counterproductive" and said AI skeptics "want to create regulatory capture under the guise of AI safety."
Yang Likun’s views are at odds with those of the “Godfather of Artificial Intelligence” Geoffrey Hinton and others. Hinton has recently supported regulation of artificial intelligence, suggesting that based on current large language models, artificial intelligence will become so advanced that it may cause harm to humans.
Yang Likun completely disagreed with this. He said that most people are too influenced by movies such as "The Terminator" and imagine that one day machines can think smarter than ordinary people. Yang Likun believes that artificial intelligence models "do not understand how the world works at all. They have no planning ability and no real reasoning ability."
What Yang Likun is referring to is general artificial intelligence. Companies like OpenAI claim that the true transformation of machine intelligence is just around the corner, but Yang Likun calls this view "overly optimistic" because artificial intelligence will require many "conceptual breakthroughs" to reach this stage. In short, what Yang Likun means is that scientists still don't know how to make artificial intelligence develop from ChatGPT to Skynet in the movie "Terminator".
However, Likun Yang did not mention the horrific biases exhibited by AI image generators, language models, or even self-driving cars in the interview. Nor did he mention how artificial intelligence could bring about a new wave of online spam and disinformation.
Yang Likun is responsible for Meta’s large language model GalacticaAI. GalacticaAI was originally designed to help researchers work more efficiently, but Meta took it offline last November after scientists discovered it cited text errors and even fabricated journals. Yang Likun expressed dissatisfaction with the removal of GalacticaAI, saying: "It is impossible to have fun by abusing robots. Is it fun to do this?"
Yang Likun also admitted that artificial intelligence will eventually become smarter, even smarter than humans. But he insists that he doesn't think AI will cause harm at all but will help us solve larger social problems, such as curing cancer or tackling climate change.
Yang Likun said: "Intelligence has nothing to do with the desire for control. If the smartest humans really wanted to dominate others, then Albert Einstein and other scientists would be rich and powerful, but they are not."
Previously, Yang Likun claimed that society has the ability to "stop evil and have unlimited power."