Now, Sora has begun to experiment with the real world......

Now, Sora has begun to experiment with the real world......

The author of this article is Yin Ye, CEO of Huada Group

On February 15, 2024 local time in the United States, OpenAI released Sora. To put it simply, it is a large video model that can be directly generated by text or language through artificial intelligence. OpenAI does not simply call it a video model, but a "world simulator". In other words, from this moment, the digital twin world that can fit more real physical laws has entered human society. I'd like to compare it to Newton's era of AI development.

In 1894, when the Beiyang Naval Division lost to the Japanese Navy, the fall of the Qing Dynasty was almost a foregone conclusion, and the main reason for this was of course the comprehensive combat quality, but in terms of the parameters of the main ships, it mainly lost in speed. The main warship of the Japanese Navy, the Yoshino, reached a speed of 23 knots, the Zhiyuan ship was 18.5 knots, and the Dingyuan and Zhenyuan two ships were only about 15 knots.

In 1945, the Yamato, Japan's largest and largest first-class battleship in human history, a monster with a full load displacement of more than 70,000 tons, was sunk after only a few battles. The US aircraft carriers encircled and suppressed them from the air, which was a dimensionality reduction attack, and it lost in the dimension of the Z-axis, and the era of battleships with large ships and huge guns came to an end.

Two years ago, I confirmed that Zuckerberg's "metaverse" was too early, but after the introduction of ChatGPT more than a year ago, I talked to many friends and believed that this wave of AI opportunities should not be missed, and we should actively embrace technological changes and increase open communication.

On the seventh day of the first lunar month, OpenAI's Sora was born, and Adobe, which is famous for its design software such as Photoshop, may never have dreamed that the stock price would be such a trend. Indeed, if a video can be generated in a single sentence, what is the future of an artist or video editor? Just as the end of instant noodles is takeout or pre-made dishes, not better instant noodles.

Friends can check out these videos, which are available on the official website. When I watched this demo video that was difficult to distinguish between real and fake, I understood more clearly that what we encountered was not a cycle, but an era. The cycle will return, but technology will only whizz by, and a real digital world that can be twinned with real laws of physics has begun its time.

Why did we all think that we could still fight when ChatGPT 3.5 was released? Because the competition at this time was still mainly at the one-dimensional level of language and text, and the arrival of Sora allowed us to see the possibility of simulating a two-dimensional and even multi-dimensional world.

Although Sora was still imperfect at this time, it was still a two-dimensional, real-world simulation. For example, what we just saw, these fake videos, if you look closely, there are a lot of flaws in them. For example, the mistake of the left leg and the right leg in the Tokyo video is very obvious. But at this time, if you look at its competitors, it is indeed a kind of Stone Age look.

A blogger named GaborCsellev on Twitter gave the four models of Sora, Pika, Runway and StableVideo, and entered the same prompt, which is the prompt, so I simply translated, the prompt he lost was, " In the beautiful and bustling city of Tokyo, where it is snowing, this shot is to walk through the busy city streets, following a few people who are enjoying the beautiful snow scene and shopping nearby, while the petals of the cherry blossoms in full bloom dance in the wind and fall in color."

We can take a look at the different generation effects of these four software, although the snow and cherry blossoms are in full bloom, in the real world, they almost do not appear at the same time. But with this simple comparison of effects, we can already see that Sora's performance is far ahead.

I'm afraid that the top student is still modest, OpenAI in its technical report, unabashedly admitted its many shortcomings, Sora in many ways can not fit the physical characteristics of the real world, but what does it matter, who would ask a toddler to just learn arithmetic, he is required to understand equations, and importantly, he is ready to continue to learn and apply calculus.

Just as the physical construction of Newton's time was not perfect, Einstein's quantum mechanics continued to supplement it, but in 1687, Newton's "Mathematical Principles of Natural Philosophy" appeared, and the scientific revolution of mankind was born. In the years before the epidemic, there were endless claims that human beings had encountered scientific and technological bottlenecks, and it seemed that we had been locked by "Zhizi", but in just four years, whether it was a breakthrough in life space-time omics technology, a breakthrough in mRNA vaccines, a breakthrough in controlled nuclear fusion, a breakthrough in superconducting materials, a breakthrough in quantum computing, and especially a breakthrough in artificial intelligence, let us once again see the possibility of the rapid evolution of human civilization.

Now, Sora has begun to experiment with the real world......

In addition to optimism, we should calmly reflect on ourselves, how much original contribution does this have from China? And in the unfavorable context of "decoupling and breaking the chain", is the opportunity and time left for us really enough? Trump began to fight the Sino-US trade war after taking office, and after a few years of fighting, he found that he injured the enemy 1,000 and damaged himself 1,000 and 2,000 himself, although Biden was constantly questioned about his physical health problems, especially brain health problems, but he had to be convinced.

Today's chip is no longer just an electronic component, a commodity, it is a provider of infrastructure, if the existing human language, images, audio, video and other databases are regarded as the gold mine of human civilization, but you already have an excavator, but I am limited to use a shovel, what will be the result in the long run?

On the issue of chips, we may have made the mistake of seeking a sword. For example, there is always a voice that thinks that we can break through 14 nanometers, or even thinks that if we can catch up to 7 nanometers, we will win, but this premise is that our competitors will not improve and will just wait in place. I'm not saying that it doesn't matter if you catch up to 14nm or 7nm, but you have to understand that competitors will achieve 3nm, 2nm, or even 1nm in the same amount of time.

We can despise it strategically, but we can't but attach importance to tactics, and the science and technology circles, the industrial circles, and the capital circles cannot pretend to understand with confusion, and we may fall into a capital routine of conceptual hype on the issue of large models. At present, according to incomplete statistics, there are more than 300 companies claiming to have large models in China, I believe that most of them are still "small and scattered", in the face of GPT in the Chinese Simplified Chinese world, it seems that it can still fight, but in the face of the emergence of Sora, can there be a fight, or even close?

For example, if you are climbing a mountain, the leader has been waiting for you on it, and finally it is almost approaching out of breath, and the leader is walking leisurely and slowly up for a while, which is psychologically and physically, constantly consuming and hitting you.

I am not talking about this paragraph to be pessimistic, let alone to grow the ambition of others and destroy my own prestige, I want to say that it is not terrible to know that backwardness, but what I am afraid of is that I know that I am backward and do not admit it, which will only get farther and farther.

In 1448, when the German craftsman Gutenberg invented his printing press, a large number of books in Europe were popularized, the exchange of academic ideas accelerated, the scope of literary works expanded, the Renaissance began to accelerate, and gradually catalyzed the scientific revolution. Similarly, telegraph, radio, television, and the Internet have greatly accelerated the symmetry of information, while artificial intelligence will promote human intelligence and knowledge equality at an unprecedented speed.

At the important juncture of carbon-silicon integration, where should our science and technology, especially our education, go? Will there be a way out and a future for blindly being a small-town problem-solver? How long will the embarrassment of Needham's question and Qian Xuesen's question last? This question is left to all the gentlemen to consider together.

Zhuangzi said in "The Theory of Things": "Zhuang Zhoumeng is Hu Die, lifelike Hu Die, self-metaphor is suitable for ambition, I don't know Zhou also...... I don't know if Zhou's dream is Hu Die, and Hu Die's dream is Zhou?" Liu Cixin's "Three-Body Problem" is the most famous sentence: "Physics does not exist." The advent of Sora has also made many people in the industry exclaim, "Reality does not exist". So, we might as well think more boldly, if we can already simulate the real world, but how do we know that the real world today is not simulated?

Now, Sora has begun to experiment with the real world......

(This article represents the author's personal views only)

Read on