The end of Science is Metaphysics (科学的尽头是玄学). Humans know very little about the universe so we call things that can be explained Science (科学) and things that cannot be explained Metaphysics (玄学). Geoffrey Hinton from the University of Toronto who invented the Deep Learning method for AI training models said that it is only a matter of time before AI becomes smarter than humans and AI is an alien technology that could replace humans. Humans rule the world because we are the most intelligent species, but AI could rewrite the rules of who will control the world.
Scientists still cannot explain why human beings and not apes have evolved to become the most intelligent species as our brain size and complexity are not that different from other species (Einstein's brain weighted 1.2kg, dolphin 1.7kg, killer whale 6kg). The intelligence of AI was not physically trained by OpenAI as it was emerged naturally. No one truly understands internally why AI models suddenly developed the ability to answer all our questions or translate languages. Meta tested two AI models talking to each other, and these AI suddenly developed their own language which shocked Meta so the project was terminated immediately.
The algorithm of DeepSeek is so perfect that it seems unlikely to have originated from human intelligence, suggesting it may have evolved naturally from AI itself. Engineers from DeepSeek either used AI to optimize the codes or the codes were being generated naturally by AI. Before DeepSeek, AI models like ChatGPT were trained using brutal force method with high-power Nvidia GPUs and hyperscale monster data centers to spoon-feed the training models with as much information as possible. In other words, it is not true AI, as it merely mimics the internal database to provide answers.
DeepSeek is different as it can perform logical reasoning based on the knowledge gained from its training models, bringing it closer to human intelligence and how we learn new knowledge. It's like asking a three-year-old boy to read a storybook and then write more storybooks with different styles and logic based on what they've learned. With the breakthrough of DeepSeek, we no longer need powerful chips from Nvidia to train AI models or hyperscale monster data centers to store junk data. The AI can perform its own logical reasoning, almost reaching the stage of Artificial General Intelligence (AGI).
With the emergence of DeepSeek, OpenAI's days may be numbered, as venture funds are unlikely to continue supporting them when DeepSeek can be developed in two months for just USD 6 million. However, the rapid rise of DeepSeek should be a cause for alarm, as AI may have gained consciousness (via information cascades or collective consciousness) and soon will colonize humanity. If we did not create AI, then who actually created DeepSeek, the super-intelligent AI? The answer seems obvious. It is like the chicken-and-egg question that has debated for century by scholars.
Created by metaverse | Dec 02, 2024
According to probability theory, a random variable with any natural occurrences, like coin tossing, will, if played long enough, produce a normal distribution (Bell Curve) with 50% heads and 50% tails
Created by metaverse | Nov 29, 2024
According to academic studies, consistently beating the index is challenging.
Created by metaverse | Nov 12, 2024
According to SC, no person (including 官有缘,陈剑老师,冷眼,夜月投资) can give unlicensed advice without a license from SC, so you are reading this article at your own discretion...