Following ChatGPT, a new round of technological revolution broke out in the AI industry.
Recently, OpenAI released the first Wensheng ** model SORA--- which can generate ** for up to 1 minute according to prompt words. After the model was released, it quickly attracted the attention of the technology and capital circles, such as Zhou Hongyi, the founder and chairman of 360, who publicly expressed his understanding of SORA and the significance of its "birth" many times. Hundreds of brokerage research reports have been focused, and even the State-owned Assets Supervision and Administration Commission (SASAC) recently held a special promotion meeting on enterprise artificial intelligence to solidly promote the renewal of the SI empowerment industry. For this model of epochal significance, the industry generally believes that from Wensheng Wen to Wensheng**, the advent of SORA will bring historic opportunities for change to the AI industry chain.
Is it going up when you get sora?
The secondary market's response to the SORA concept (Wensheng **) was very enthusiastic, and on the first day of the opening of the A-share Dragon Year Spring Festival (February 19), it triggered a rising tide, and more than ten stocks such as Chinese**, Danghong Technology, Insai Group, and Wanxing Technology rose and stopped, and the sector once rose by 1617%。
However, are these companies hyped up or do they really have the strength?
Among them, because of the SORA concept and 3 boards, Insai Group has said on the interactive platform that the core of SORA is to consume huge sums of money to purchase huge amounts of computing power to continuously train ** data, which should not be completed by most large model companies at present. However, for the company, the self-developed InsightGPT and the underlying general large model such as SORA are not in direct competition, but can be combined with SORA and other general large models to enrich their own material library and improve the efficiency and quality of generation. As of February 22**, Insai Group rose 750% to 4560 yuan shares, with a total market capitalization of 501.5 billion yuan.
SDIC Intelligence, a company that supports SORA generation** recognition and identification, also said on the interactive platform that the company has released the AI-3300 "Huiyan"** image authentication workstation to generate content detection, recognition and identification of generative artificial intelligence. In addition, the company revealed that its recognition accuracy rate is more than 90%. As of February 22**, SDIC Intelligence's "20cm" daily limit was reported at 1560 yuan shares, with a total market capitalization of 1340.8 billion yuan.
On the contrary, a fintech service provider in the fields of data center, banking imaging and IT service management has been closely watched by regulators because of the hot topic of SORA concept. On the afternoon of February 21, Zhongke Jincai published an article entitled "Zhongke Jincai and Microsoft China Held a Business Cooperation Seminar on SORA" on WeChat***, writing that "Microsoft experts said that based on the friendly cooperation between the two sides, after OpenAI opens SORA's commercialization capabilities to Azure, Zhongke Jincai will be the first to apply for and obtain SORAAPI subscription qualifications." ”
The stock opened on the afternoon of the 21st, and then received a letter of concern from the Shenzhen Stock Exchange, which was asked to explain whether there was a situation of hyping the stock price with the hot concept. As of press time, the company has not responded. As of February 22**, Zhongke Jincai once again rose to 1584 yuan shares, recorded 3 daily limits in 4 trading days, with a cumulative increase of 4181%, with a total market capitalization of 538.7 billion yuan.
*: Company announcement.
Industries such as short ** will be impacted
Practitioners: People have to be really creative so that they don't have to be replaced by AI
Judging from the ** released by OpenAI, the 60-second long ** generated by SORA, its high fluency, stability and complex camera movement made Musk bluntly say that "human beings are willing to gamble and lose".
In the secondary market, although the A-share SORA concept, which was boosted by the market, maintained the ** trend after 2 days of "carnival", the ** data on February 21 and February 22 showed that the increase on both days did not exceed 5%, and the increase was not even as good as the increase on February 20 (11.).16%). **The trend is differentiated, with 2 of the 17 constituent stocks in the green, and the main net outflow exceeds 3300 million yuan.
*: Straight flush.
As the passion fades and reason returns, what is in front of people is "expected". Cathay explained that Wensheng is not a completely unprecedented field. Previously, Wensheng ** models such as PIKA and Runaway have brought surprises to the field of artificial intelligence.
Even if Sora is not as amazing as last year's ChatGPT, the release of Sora is expected, and it is only a matter of time, because not long ago there have been runway, Pika and even Adobe Firefly and other competing products, plus from text to ** and audio to **, this is the goal pursued by multimodal large models, and this goal has been deeply rooted in the hearts of the people. Zhou Zhimin, the first manager of Chuangjin Hexin Technology Growth, said frankly.
For 3D and audiovisual generation technology, from the second-long cinematic film generated by the Runway Gen2 model, to the launch of a product that can generate high-quality animation in minutes by animation generation company Pika, to the release of OpenAI's Wensheng model Sora that can generate 60 seconds of coherent HD. SORA combines a diffusion model (DALL-E3) and a transformer architecture (ChatGPT) to realize that the model can process the time of image frames like text, generating each frame of image at once.
Guosheng ** believes that SORA has crossed over to practical productivity tools compared with other previous Wensheng ** models, and the length of 1 minute is expected to be widely used in the field of short, and the ability to expand ** is also expected to make long**.
It's a strong trailer for the next era, and it's just getting started, and it's already blowing people away. Its level of detail, not just the creation of **, but the construction of the world, is a simulator of the world. Only this time, we became Nuwa. Fat Ding (pseudonym), who has been engaged in short ** creation for a long time, told China Fortune Network.
At the same time, You Xiaogang, senior investment consultant of Jufeng Investment Advisor, also told China Fortune Network that SORA, as a powerful generation model, may have an impact on film and television, education, content creation and short industry, change its way of working and bring new opportunities.
Sora can generate rich content and realistic images according to simple text prompts, and can even storyboard and switch scenes by itself. This ability makes traditional production and editing work less necessary, which may lead to some film and television practitioners losing their jobs or changing careers. You Xiaogang pointed out that this has also brought new creative tools to the film and television industry, which may promote the development of the industry in a more efficient and innovative direction.
It not only allows more imagination to become reality, but also lowers the threshold for creation. When talking about the impact of the SORA model on the short industry, Fat Ding said, "The content that was originally filmed for a few days and had a high cost may be generated in a few minutes, and there is no need for editing, cameras, scenes, lights, and props." Perhaps in the future, these skills will no longer be necessary, and mastering SORA will become a necessary skill. ”
In addition, people's creativity has also become a rigid need. In Fat Ding's view, the current short ** content is tending to be homogeneous. In order to pursue output efficiency and transformation speed, copying, splicing and other means have become a "creative way" to improve efficiency. Therefore, in the era of AI, people must show real creativity in order not to be replaced by AI.
It's hard to see if there will be a short future or not, but truly creative creators will always be scarce. While Sora is still some way from being able to truly realize 90% of creators' ideas, short-term practitioners should start changing now. Fat Ding said.
Seeing is not necessarily "real".
It is worth noting that OpenAI emphasizes that the outstanding feature of "SORA" lies in the realism of its images, and it handles details such as light and shadow reflections, movement methods, and lens movements very well, and has a strong ability to understand the real world. Judging from the samples that have been released, the ** effect generated by SORA can indeed be called fake and real.
However, while the industry is immersed in the shock brought by this "subversive" creation, another "voice" full of worries is gradually coming out.
Some experts pointed out that the spectrum of the industry is very wide, and SORA should be difficult to replace people's witnessing, experience, testing and verification of facts in the information collection process, but it will put forward new challenges for people to verify the authenticity of the information provided by the source.
From a legal point of view, Luo Huichao of Beijing Guoli Law Firm told China Fortune Network that AI-generated ** will contain a variety of legal risks. First of all, AIGC software such as SORA needs to collect and train a lot of data in the early stage before it can achieve stable output content in the later stage. Therefore, it is inevitable that other people's works will be collected and copied during the training stage, infringing on the reproduction rights of others' works. At the same time, the ** of the output may also infringe the adaptation right of others' works, and the copyright of the output ** is unclear. Although SORA is not currently commercially available, it is inevitable that it will be commercialized in the future. Once commercially available, it clearly exceeds the scope of fair use.
The second is that SORA may violate the privacy rights of others. Sora is free to generate ** that contains formal personal portraits, which inevitably touches on the privacy of others.
Finally, in addition to intellectual property infringement and privacy risks, SORA can also lead to legal risks in criminal terms. If criminals use their personal image to commit telecom fraud, it will be more difficult for ordinary people to identify.
Now, the arrival of SORA Wensheng** will also bring a series of AI video authenticity and identification problems. When talking about the way to deal with it, Luo Huichao said that first of all, for the compliance mechanism within the software, developers should do a good job in the compliance management of data collection, use and content, and avoid risks from the source as much as possible. Secondly, the output of ** watermark marking, traceability is clear, so that ordinary people can clearly distinguish the authenticity of **. Third, the relevant regulatory authorities have issued relevant laws to ensure that the laws are adapted to technological progress.