Shulou(Shulou.com)11/24 Report--

Thanks to CTOnews.com netizens Xiao Zhan cut, South China Daniel Wu's clue delivery! CTOnews.com, June 16, according to CNBC in the United States, Yann LeCun, Meta's chief artificial intelligence (AI) scientist, said at the Viva Tech conference in Paris yesterday that AI is not as intelligent as dogs and should not be regarded as a threat to humans.

▲ source Pexels when asked about the current limitations of AI, Yang Likun said: "those AI systems are still very limited, they do not understand the basic situation of the real world, they just train based on a large number of texts. But most human knowledge has nothing to do with language, so this part of human experience will not be captured by artificial intelligence."

Yang Likun also said, for example, that AI can now pass the US bar exam, but cannot install a dishwasher, but a 10-year-old can "learn it in 10 minutes."

In response, Yang Likun concluded: "what this case tells you is that we [the current technological development] are missing out on something very important. AI should not only reach the level of human intelligence, but even reach the level of dog intelligence."

Yang Likun said that Meta is working on training AI based on video content, which is more difficult than language training. As for the significance of this kind of training, he said, for example, a five-month-old baby will not think too much when he sees an object floating in the air, but a nine-month-old baby will already be surprised by this. Because he has realized that objects should not normally be floating.

Yang Likun believes that machines smarter than humans will emerge in the future, but they should not be seen as dangerous: "We should regard it as a very useful thing. Each of us will have an AI assistant who will help you in your daily life like an entourage and smarter than yourself." He said the AI needed to be designed to be "controllable and basically subordinate to humans", and he refuted the idea that robots would take over the world.

CTOnews.com previously reported that Meta recently released its first "humanoid" AI model, I-JEPA, which can analyze and complete missing images more accurately than existing models, which is Yang Likun's ideal prototype of truly intelligent AI.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.