Reading guideWith the rapid development of science and technology, AI technology continues to evolve, from machine learning to neural networks, to Transformer models, especially the stunning debut of ChatGPT in 2022 and the emergence of GPT-4 in 2023, pushing AI development to new heights. On February 16, 2024, OpenAI released the latest "Wensheng**" large model SORA, which has aroused widespread public attention to generative AI. The model is a diffusion model based on the Transform architecture, inheriting the image quality and instruction-following ability of DALL·E3, and using GPT expansion, which has three advantages: ultra-long generation time (60s), single-** multi-angle lens, and understanding the physical world. Not only that, but it can also take existing static images and generate them from them, as well as take existing ones to expand or fill in missing content. On February 19, the Spring Festival holiday ended, and A-shares ushered in the first trading day of the Year of the Dragon0. ChatGPT, virtual humans, and AIGC all have **. All this indicates that the industrial revolution of creative content that "everything can be generated" has arrived, and the future world will be full of infinite possibilities and surprises. In the third issue of GT Technology Watch, we tried to focus on the following three topics:1.AIGC Focus: Content can be generated2.What industries will SORA disrupt? What are the new industrial opportunities behind it?
3.Will machines replace humans? How to effectively deal with the anxiety caused by AI?
01.AIGC Focus: Content can be generated
Artificial intelligence can be divided into decision-making AI and generative AI according to its model. Decision-making AI mimics the human decision-making process, and its application fields include face recognition, recommendation systems, risk control systems, other intelligent decision-making systems, robots, and autonomous driving. Generative AI focuses on the creation of new content, by learning the joint probability distribution in the data, summarizing the existing data, and using deep learning technology on this basis, to create imitation and stitching content, which is equivalent to automatically generating new content. In 2021, Gartner Consulting** found that 20% of content will be created by generative AI by 2023, and by 2025, generative AI will generate 10% of all data (up from less than 1% in 2021). In September 2022, Sequoia Capital's official website published an article titled "Generative AI: A New World of Creativity": Generative AI will generate trillions of dollars in economic value. According to **, in 2025, the scale of domestic generative AI applications is expected to exceed 200 billion yuan, and the application space in the domestic media field will exceed 100 billion yuan, accounting for half of the country. Generative AI can generate content in a variety of forms, including text, audio, and more. In text generation, generative AI uses language models, neural networks, and deep learning technologies to quickly create a large amount of content that can help improve customer experience, such as news, scripts, marketing texts, and intelligent customer service. In generation, generative AI analyzes through computer vision to generate marketing materials, design plans, and artworks. In audio generation, generative AI helps users better analyze, edit, and generate audio files, thereby helping to create excellent audio works, such as cloning real people's voices, generating specific voices from text, composing and arranging, etc. In Generation, it can help users generate high-quality products, such as detecting and deleting specific clips, tracking clips, generating special effects, compositing, etc. The AI digital human of the fire is also its "specialty". In cross-modal generation, generative AI can generate ideas from text, generate from text, generate from text, generate text from or generate text. In games, generative AI is used in game development to automate game design and enable better gaming experiences, such as AI NPCs.
02.What industries will SORA disrupt? What are the new industrial opportunities behind it? The advent of SORA has sparked heated discussions in the tech community. Tesla founder Elon Musk posted a news release: "Humans enhanced by artificial intelligence will create the most outstanding works in the next few years." "Generative AI has a wide range of application scenarios, not only in text, audio, games and other digital, but also in manufacturing, education, healthcare, film and television culture and other physical industries. Schools can use AI generation technology to make teaching, hospitals can use AI generation technology to simulate the surgical process, games, virtual reality, meetings, etc. may be related to AI generation technology....Many emerging technology companies use AI technology for film and television production, combining traditional film and television production methods with artificial intelligence technology. Yuan Pengfei, CTO and co-founder of China Film, said that the launch of SORA has a far-reaching impact on the film and television industry, and the production process model of the film and television industry will be redefined and rewritten. First, SORA, as a representative of the wide application of AI technology in the film and television industry, may become a new digital content production model, changing the original pattern of film and television, advertising, animation and other first-class content, opening a new chapter in film and television production, and bringing new horizons and possibilities to the film and television industry. Second, because SORA can directly generate ** according to the input text, it eliminates the cumbersome production process of the traditional **, greatly improves the production time of the traditional **, and is expected to reduce costs and increase efficiency for enterprises in related industries. In general, SORA, as an advanced AI tool, has played a role in promoting the development of the film and television industry, and has also brought some new opportunities to individual creators. The development of AI models is accelerating the arrival of AGI general artificial intelligence, and throughout the entire AI industry chain, the cooperation of hardware and software can better realize the value implementation of a variety of terminal applications. The launch of the SORA model will also directly promote the development of AI model production tools (AI algorithm framework, AI development platform, AI open platform), AI computing power foundation (AI chips, servers, intelligent computing centers, cloud services), and AI data resources (AI basic data services, data governance) in the upstream of the AI industry chain. For downstream applications, the positive impact is all-encompassing. At the hardware level, it will directly promote the development of AI servers, optical communications, switches and other industrial chains.
Will machines replace humans? How to effectively deal with the anxiety caused by AI?
In the face of the problem of humans and machines, everyone expects the computers of the future to do more things, so much so that some people wonder: will there be anything to do on their own in 30 years? Generative AI is powerful and has a wide range of applications, and practitioners related to text and audio content will also feel anxious and afraid of being replaced by it when they face the launch of the "powerful assistant". Will machines replace humans? Venture capitalist Mark Anderson categorically declares that "software is eating the world," while his colleague Andy Kessler seems happy when he explains that the best way to increase productivity is to "get rid of humans." In this regard, Peter Thiel, founder of PayPal and Facebook's first external investor, has said that computers are tools to assist humans, not substitutes. In the coming decades, the most valuable industries will be built by entrepreneurs who develop computers to empower humanity, not to eliminate it. Technology means complementing. There is a fundamental difference between what humans and machines are good at, and humans are conscious and good at making plans and decisions in complex situations, but they are not good at processing large amounts of data. Computers, on the other hand, are good at efficient data processing, but they can't make basic judgments that humans can easily make. The stark difference between humans and machines means that the results of working with computers are much higher than those of dealing with people. Yuan Pengfei believes that at this stage, SORA can be understood as an intelligent creation tool, which does play a role in efficient production in the fields of film and television preview previz, film and television special effects, personal creative short, material generation, etc., but in commercial projects, such as film and television dramas, advertising, three-dimensional animation, short ** and other fields, it is difficult to achieve commercialization in the short term, and it is difficult to meet the needs of planning, directors and other content customization. Most of them are realistic styles, and there is a lack of more artistic styles, so it is difficult for SORA to meet the needs of commercial projects such as customization, stylization, and uniqueness, and the generation is uncontrollable, and there will inevitably be a large number of repeated debugging and modification, and it is impossible to replace the existing film and television production process. How to effectively deal with the anxiety caused by AI? Fu Sheng, an entrepreneur of AI model application, shared in CCTV Finance Guanghualu that the first thing to do is to change the mode of thinking and embrace new technologies. Don't think of AI as a supplement, but rather a deep understanding of the underlying changes it brings. AI is essentially an affirmative action tool, not exclusive to a few people, and for ordinary people, it is believed that it can bring rights and interests to themselves. Second, pay careful attention and learn. The SORA model still has shortcomings, such as the difficulty in accurately simulating complex scenarios, and the inability to understand some specific causal and temporal connections. In addition, the model may also confuse spatial details in some text instructions, such as left and right directions. Going forward, human-machine solutions will be able to bridge these gaps and drive technology in more areas. With the release of the SORA model, AI technology is once again at the forefront of technology, and the new wave it leads not only heralds disruptive changes in the industry, but also brings unprecedented industrial opportunities. However, every technological leap comes with societal doubts and anxieties. In the face of the changes brought about by SORA, keep an open mind, look at the development of AI technology rationally, resolve anxiety with wisdom and courage, welcome the arrival of a new era of AI, and explore the future of infinite possibilities.