The most troublesome giant in 2024 is still OpenAI.
First, the founder and CEO Altman, revealed that he expects to raise $7 trillion to improve global chip manufacturing capacity.
7 trillion US dollars, can be bought by Apple, Microsoft, and Google together.
Compared to the astronomical figures, what is even more shocking is another thing about OpenAI.
In the early hours of February 16, its firstWensheng ** model soraOfficially unveiled, and released several AI-generated **, up to 60 seconds.
AI**, since then it has reached the level of being fake.
Seeing these, ordinary people shouted "reality, it doesn't exist anymore", and their peers all withered.
Stability AI, a star AI company, originally released a new **model, but after seeing Sora's**, the official quickly deleted its own tweets.
The founder of another company, Runway, lamented that it used to take a year of work, but now it is done in a few days and hours.
Musk left messages everywhere "Humans are willing to gamble and lose (GG Humans)".
CCTV was also alarmed, and multiple channels reported on SORA around the clock.
For Sora, there are only two kinds of people who are happy in the whole process.
One is the old color batch, they have fast-forwarded to the association of "MR headset + AI creation plot".
One is naturally OpenAI, whose valuation reached $80 billion and may soar after driving its peers crazy again.
Sora can only generate 60 seconds**, why is it so valuable?
2 hours after release, so that the opponent lost hundreds of millions
In the field of AI generation, OpenAI is not the first to eat crabs.
The difference is that the previous AI ** companies, the AI generation they did was basically only a few seconds, and it was vague, and there was an obvious sense of disobedience.
For example, last year, an AI-generated ** of a foreign male star eating pasta was fake at a glance.
So many bigwigs say that we still have a long way to go in terms of AI generation.
But its commercial value cannot be denied.
Previously, the most well-known runway in the industry could only generate ** for 18 seconds at most, but with this skill, runway participated in the production of the Oscar-winning film "Instant Universe", and the valuation once reached $1.5 billion.
There is also a start-up company PIKA, which was established at the end of last year, which became a sensation with its Wensheng ** model. The company, which has only 4 people, completed three rounds of financing in half a year, and its valuation soared to 2$500 million.
I just didn't expect thatSora, who was born out of nowhere, destroyed everyone's dreams overnight.
The CEO of Runway immediately said two words on the Internet: "Game on."”
The CEO of Stability, another AI company, bowed down: "Ultraman is really a magician. ”
Shutterstock, the largest merchant in the United States, was hit across the border, and its market value lost about $70 million in 2 hours after Sora was released.
What's even worse is Google, Wang Feng in the technology industry in seconds.
Because on the same day that Sora was released, Google also released a new generation of multimodal large model Gemini 15 Pro, a complete spike in OpenAI's GPT-4 Turbo in terms of performance.
It was supposed to be a day for Google to raise its eyebrows, but it was stolen by Sora that made all the headlines.
To overshadow all companies, SORA is too subversive.
Others are still using AI to painstakingly generate 4 seconds**, and Sora is randomly generating 60 seconds of HD blockbusters.
Other AI-generated** are full of copycat feel, and Sora can even generate movie-level scenes to make sense of reality.
Even the multi-camera in the "same scene"**, sora has been sorted out.
Whether it is the best duration, picture fineness, or detail integrity, whether it is a shot to the end or a seamless switch between multiple cameras, SORA's performance can be described as crushing opponents.
Some people use the same prompt words to compare the ** generated by SORA and other similar products, and the difference is clear at a glance.
What's more, the barrier to entry for using SORA is extremely low.
Judging from the official publicity**, you only need to enter prompts such as "constituent elements, environment, behavior and order of occurrence, **style", and you can generate 60 seconds of high-definition, coherent, and rich camera movements and transitions in different styles**.
So, it's also easy to use it to generate other styles of virtual **, such as 3D or anime.
At the same time, it can also directly generate ** of different sizes according to the requirements and adapt to the screens of different devices.
Maybe soon, everyone can have a director's dream.
It is no secret that SORA is worth hundreds of billions of dollars
In fact, SORA still has many obvious weaknesses.
OpenAI notes that it may struggle to accurately simulate the physics of complex scenarios and may not be able to understand cause and effect.
For example, in this **, before and after the old man blows out the candle, the flame does not change in the slightest, which is unreasonable.
The glass shattered, but the relationship between the flow of the liquid and the cracking of the glass was also unreasonable.
A person running upside down on a treadmill is even more unreasonable.
For AI, no matter how simple and natural a phenomenon is in the real world, it may need to be learned to understand and restore.
"AI painting", which was very popular last year, had a similar problem at the beginning.
Let it draw a picture of "sleeping on the ground", which may show "floor + pillow + human head", thinking that putting your head on the pillow is sleeping.
It has also been discovered that AI painting does not always allow the characters to eat ramen honestly. What tools to use to eat noodles, how many sticks to eat in one bite, AI needs to learn.
However, AI painting technology quickly iterated and became a work tool instead of being mentally retarded.
A person who works in illustration design said that after using AI painting, the work is more efficient, "Now Party A gives me a theme, and I can generate seven or eight pictures for him at the same time." ”
Now there is also an obvious hard sora, and it also has the ability to iterate quickly.
This is mainly due to its ability to understand long texts, which is also OpenAI's original killer feature.
For example, on the input side, OpenAI uses GPT to first accurately and exhaustively expand the prompt words entered by the user, and then hand over the expanded prompt words to SORA, so that SORA can better follow the prompt words to generate accurate **.
Therefore, compared with the previous AI ** generation, you can only rigidly translate the copy, and then generate a rigid **, SORA can not only understand the copy, but also play on its own based on the copy, making the ** more natural and exciting.
Just like this**, on the premise of completing the theme, Sora automatically made multiple lens transitions.
With better comprehension, there can be better logical ability.
Therefore, after understanding and mastering more real-world phenomena and laws, the ** generated by SORA will get closer and closer to reality.
The window of the moving train is occasionally blocked, and the brief appearance of the reflection of the characters in the car is very amazing.
You can also have a movie trailer with the texture of a Hollywood blockbuster.
In the close-up view, the lens is full of detail.
That's why OpenAI defines SORA as:"World Simulator".
In addition to being close to reality, SORA can also realize the unrealistic brain hole of human beings. For example, based on the prompt "realistic close proximity of two pirate ships fighting in one cup**" it generates such a **.
No wonder many netizens are saying that they are going to be unemployed, and many industries such as film and television, advertising, and games will be impacted.
Many people also believe that even short ** platforms such as Douyin and TikTok should have a sense of crisis.
After all, once AI generation is on the right track, the output is endless, and AI can better understand the mechanism of human addiction, if OpenAI launches its own short** platform, it is likely to become a hegemon.
OpenAI's possible ambitions are getting more and more scary.
From huge losses to wild gains,
Why has it become a nightmare for giants?
As the top of the technology industry in the past two years, for a while, OpenAI was not optimistic.
In the first few years of its establishment, OpenAI had no profitability and relied on its father to feed itself.
So much so that there have been business reports that OpenAI costs about $700,000 a day just to run its artificial intelligence service ChatGPT, and OpenAI is currently in a state of burning money, and if it does not accelerate its own commercialization process, it is likely to have to file for bankruptcy at the end of 2024.
But OpenAI said with action: It's not that it won't make money, it's just that it doesn't want to make money yet.
Since the second half of last year, OpenAI has accelerated its commercialization. According to the founders, as of November last year, 92% of Fortune 500 companies were using OpenAI's products. Throughout 2023, OpenAI's revenue exceeded $2 billion.
At the same time, its most famous product, ChatGPT, has reached 100 million weekly users.
With that, its valuation, which has skyrocketed from $30 billion at the beginning of last year to $80 billion today, is second only to ByteDance and SpaceX among startups.
However, compared with the earning power and soaring valuation, OpenAI has more feared characteristics for giants.
That is the ability to create an outlet.
In the past 14 months, OpenAI has always been at the top of the wind.
From AI chatting, to AI painting, to AI generation**, OpenAI's every appearance seems to be a slap in the face of giants.
As soon as ChatGPT debuted, Google was scared to urgently launch the AI-blessed search engine Bard, but the result was unfavorable, and the stock price fell 74%, and the market value evaporated by $102.4 billion.
Sora just sent out a few paragraphs of AI**, stealing all the headlines of the technology section, and embarrassing his peers who have been working hard for a long time. Zhou Hongyi also posted a long post on Weibo saying that SORA means that the implementation of AGI will be shortened from 10 years to 1 year.
Today's hot generative AI market can be said to be inseparable from OpenAI's step-by-step boost.
A start-up company, preemptively creating a trend, enlarging the market, and faintly showing signs of becoming a new giant, is enough to become a nightmare for all the current giants.
But instead of looking at OpenAI to continue to create outlets, what the giants should do is to reflect.
Wang Xiaochuan, former CEO of Sogou, once commented: "The success of OpenAI is, first of all, the victory of technological idealism. China needs its own openAI, and it needs technological idealism. Large factories are driven by their own businesses, and start-ups chasing capital hotspots are even more deformed. ”
In fact, this problem exists both at home and abroad.
In the process of development, the importance of technology and innovation has unconsciously taken a back seat, and all actions have become the service of income.
OpenAI is like a shadow of the past, reminding them that they must remember their long-lost hard work.
From ChatGPT to Sora, in the face of the impact of AI, ordinary people in many industries have begun to have a sense of crisis, and giants should not be delusional to continue to lie flat.
On the Internet, please contact to delete the infringement.