At a new stage in his career, Yang Likun, former chief AI scientist at Meta, believes that setting up a “new laboratory” — that is, a startup that takes into account basic research — is currently the new track with the greatest potential for development. The new model architecture he proposed uses video data to help artificial intelligence models understand the physical laws of the real world, so that they can more accurately predict future events. The model will also rely on “emotion-like mechanisms” — that is, past experiences and evaluation results — to guide the prediction process. “If I pinch you, you'll feel pain. This in turn affects your cognitive model of me. By the next time I raise my hand close to you, you'll instinctively back off. This is your predictive mechanism at work, and the 'emotion' that triggers this response is fear or avoidance of pain.” he explained. Yang Likun said that within the next 12 months, people will be able to see a “prototype version” of this technology, and larger scale applications will be implemented within a few years. Although this is far from being a true superintelligence, it has paved a viable path for it. “Maybe we're still unaware of some potential barriers, but at least we can look forward to the future.”

At a new stage in his career, Yang Likun, former chief AI scientist at Meta, believes that setting up a “new laboratory” — that is, a startup that takes into account basic research — is currently the new track with the greatest potential for development. The new model architecture he proposed uses video data to help artificial intelligence models understand the physical laws of the real world, so that they can more accurately predict future events. The model will also rely on “emotion-like mechanisms” — that is, past experiences and evaluation results — to guide the prediction process. “If I pinch you, you'll feel pain. This in turn affects your cognitive model of me. By the next time I raise my hand close to you, you'll instinctively back off. This is your predictive mechanism at work, and the 'emotion' that triggers this response is fear or avoidance of pain.” he explained. Yang Likun said that within the next 12 months, people will be able to see a “prototype version” of this technology, and larger scale applications will be implemented within a few years. Although this is far from being a true superintelligence, it has paved a viable path for it. “Maybe we're still unaware of some potential barriers, but at least we can look forward to the future.”