OpenAI's Sora, can China not keep up?

During the Spring Festival, OpenAI’s Sora became popular, and I also took a look at it. To be honest, the shock to me this time was not as good as when GPT4 came out.

I am a programmer. For me, this text-generated video function is not practical. I would rather see GPT’s programming capabilities strengthened.

Even apart from the programmer community, Sora’s uses have been overly exaggerated by some self-media that generate traffic.

There is even a weird thing like Sora course appearing before Sora is actually online. The scythe is raised high, ready to start cutting.

Vincent Tu has been highly hyped in the past year, but the problem of style consistency between AI-generated pictures has not been effectively solved, and it is still difficult to produce long-form comics with consistent style and high quality.

It is unrealistic to splice multiple 1-minute videos into a movie because the movie also contains complex plots, characters and dialogues.

Some people believe that Sora can help create scenes in the metaverse and build a virtual world for people to enjoy. But the metaverse requires three-dimensionalization, human-computer interaction and programming, and Sora is still a long way from realizing the ideal virtual world in "Ready Player One".

In "Ready Player One", a helmet is a necessity, and Apple's latest Vision Pro VR device has attracted much attention. However, there have been recent reports that some fans are dissatisfied with this 600-gram helmet, complaining that it is uncomfortable to wear and can easily cause headaches and eye fatigue. As a result, some fans have requested returns.

Sora’s impact on short video creators will be very significant. If Sora’s commercial fees are similar to GPT-4 ($20 per month), this will greatly reduce the cost of short video production. The threshold is expected to trigger a wave of explosive growth.

But don’t forget, with the equalization of technology, the essence remains unchanged. The core is still creativity, creativity, creativity.

We always overestimate the short-term impact of a new technology and underestimate its long-term impact. Just like last year’s GPT4, when it first came out, everyone thought it was over and programmers would be replaced. .

This year has passed. In fact, it is not what everyone thought. Using GPT4, programming efficiency has indeed improved tremendously, but as I said in As written in the article , GPT4 can generate 95% of the code, but it still needs the guidance of the programmer. It doesn’t know if there is an error, and it still needs the programmer to come. To identify , it will not do requirements analysis, nor will it effectively split requirements into modules, all of which require programmers.

In the long run, CRUD programmers will definitely be eliminated. OpenAI will force programmers to move up, from coding to detailed design, architecture design, demand analysis, and customer communication. ... Until true general artificial intelligence emerges and completely wipes out this industry, by that time, most existing industries will be wiped out.

In addition, everyone must remember that every time a technology company launches a new product, it will create a beautiful package. In the GPT4 demonstration video last year, Socratic teaching method was used to teach The students' solution of linear equations in two variables is very shocking, and it feels like the education industry is going to be subverted.

But in fact, solving a linear equation in two variables is just a very beautiful example, which was carefully selected. When you really use this method to learn new and complex knowledge, you will find that it is not the case at all. It is difficult for GPT4 to teach you effectively and step by step.

I believe that the video released by Sora this time is also carefully selected to make people think it is very awesome. If it is actually used, it will definitely be inferior.

Last year, many large-scale model companies in China boasted that they had caught up with or surpassed OpenAI. I feel that they were only close to the level of GPT3.5.

"There is only one kind of artificial intelligence in the world: GPT4." I have been saying this for a year, but there is still no one that has surpassed it in China.

Now that Sora is out, the gap between domestic AI and the United States has not narrowed, but has further widened.

This gave me an ominous premonition: in the AI ​​era, will the United States once again occupy the foundation, while China will use its huge market advantage to toss and involuntarily roll in at the application layer? Just like in the Internet age, it's really sad.

Tongyi Qianwen, Xinghuo, Wen Xinyiyan, Baichuan...you have to work hard! You must keep up!

The above is the detailed content of OpenAI's Sora, can China not keep up?. For more information, please follow other related articles on the PHP Chinese website!