Since the United States OpenAI released the SORA model, the Internet is boiling, saying that the gap between China and the United States AI has widened, from education to culture to system, all kinds of demeaning Chinese technology.
Last time I said that the sora model was "carved on", it was not very useful, and some people were not convinced, they said that when the "steam engine", "internal combustion engine", and "electricity" were first invented, they were not as ......effective as horse-drawn carriages
Yes, but don't forget that these things are closely related to the real physical world, and the steam engine was quickly used in textiles and navigation, the internal combustion engine was quickly used in transportation, and electricity was quickly used in lighting and power systems.
All the "industrial revolution" and "scientific and technological revolution" in history are closely related to the reality of production and life, in short, they are all "improving productivity", and the actual performance is steel production, power generation, and communication engineering.
Some people mention computers and the Internet, very well, when computers were first invented and applied, they were designed and manufactured for lethal purposes, such as missiles and nuclear bombs; The Internet was originally the ...... of the Pentagon's internal means of communicationTo put it bluntly, it is still the level of the electronics industry, computing power and communication efficiency, and the purpose is to improve "productivity" or even "killing efficiency".
Some people say that the AI of China and the United States has opened up the gap, I don't quite understand, AI is also divided into several types, some AI is responsible for optimizing industrial processes, some AI is responsible for autonomous driving, some are responsible for meteorological monitoring, some AI is responsible for commanding swarm unmanned equipment, some AI is responsible for generating bullshit nonsense, and some AI is responsible for generating***In some aspects, China has an advantage.
For example, the domestic "industry model" has been applied to mining, meteorology, medicine, molecule, electric power, waves, finance and other industries, in the mining scene, the underground scene can be transmitted to the surface through the panoramic puzzle technology of 5G + AI on the mining scene, and the mining machine can be controlled on the surface; It can detect foreign objects in the main transport belt, and the recognition accuracy reaches 98%; It can monitor the underground operation specifications and give early warning of illegal actions. In the railway inspection, it can help inspectors to automatically evaluate the image quality, locate and identify the fault of small samples, etc., and in the Zhengzhou railway section test, the accuracy of defect and fault monitoring reaches more than 99%.
The domestic meteorological model is the world's first AI model with an accuracy that exceeds the traditional forecasting method, which can complete the global weather forecast for the next 1 hour to 7 days in seconds, and the accuracy exceeds that of traditional digital analysis methods while increasing the speed by 10,000 times.
Most of the domestic AI models are closely related to the "physical world", such as meteorology, oceanography, agriculture, earth, machinery, aviation, aerospace, civil engineering, geology, etc., and solve problems in the real field.
In these areas, they are no worse than the United States, and even ahead.
Just ask the simplest question - the United States AI is so powerful, how many unmanned docks and unmanned factories does it have?The container throughput of Shanghai Port has ranked first in the world for 14 consecutive years, and its volume has reached one-tenth of the annual throughput of global ports, which is equivalent to the combined throughput of all ports in the United States ......It is precisely because of the use of a large number of unmanned intelligent technologies.
It is not that the United States has not used AI in the "physical world", they have used AI to simulate wind tunnels, and as a result, new aircraft test flights fall every day, and hypersonic missile test launches are ......crooked every day
We do have to work on many AI fields, such as protein design, industrial process optimization, and assisted recognition and judgment ......But the road of OpenAI may not be the ultimate development direction of AI, ChatGPT can make a piece of bread no matter how much it talks about lotus?Sora is full of tricks, can he conjure up an aircraft carrier?
Many half-witted people say GPT and SORA as "productivity improvement", I will ask, is it you who make you water articles and do ** at a higher level?Or will it make you more money?Judging from their principle, what is "fed" with garbage data will eventually generate garbage, and bells and whistles garbage is also garbage.
The principle of SORA is not complicated, and there is no essential difference with text generation**, **is a single frame**, relying on "power bricks flying", how can there be any "understanding of the world" that blows amazingly. Algorithms are only one aspect, and a large amount of underlying data is needed to "feed, a large number of manpower for data cleaning and manual annotation, in the final analysis, it is still up to manpower to pile up, do you say this is "smart" or "not intelligent"?
You regard SORA's performance as "leading technology", but it is better to see it as a PPT show with strong computing power and high investment, which attracts the carnival of the financial investment circle.
The direction of OpenAI's technology development requires huge computing resources and extremely high energy consumption. Sam Altman believes that AI will consume far more electricity than humans would ......expectHe even thinks that this technical route cannot be truly realized until controlled nuclear fusion is realized, so can it be considered that his route itself cannot be realized?
Yann Lecun, a Turing Award winner and one of the Big Three of Deep Learning, commented: "ChatGPT doesn't really understand the real world, it's just pure text training, and most of human knowledge has nothing to do with text." As for Sora, he said: "The process of generating ** is completely different from the causality** based on the world model". Translation translation - the one generated by the sora is not the ** that the creator wants at all, but the xjb generation.
Sora doesn't understand the "world" at all, and what is the so-called "intelligence" of something without logic and causality?
How many times has this been?
Blockchain, Bitcoin, Metaverse, ChatGPT, ......Wait until Sora attacks, and everything will be fine?
It's not that I don't like sora.,On the contrary, I'm also looking forward to making a ** of "Hawking in Lori Island".,Sora is good.,But it's not so much that many people will blow it as a "dimensionality reduction blow".
American scientific research is moving away from reality to fiction, those well-known institutions and companies are pouring a steady stream of resources into some bottomless pit and the emperor's new clothes, capital is frantically hyping ethereal concepts that cannot be implemented, and the smartest brains are studying things that have nothing to do with the real physical world, and Bo is a "miracle" that is impossible to achieve.
Don't you think it's ridiculous in the old age of the empire? If you don't study steel and energy, don't study the liberation and development of productive forces, don't study how to make everyone's life better, and focus on "cyber bliss"?