The project I worked on the most with ChatGPT was writing scripts. ChatGPT is more patient than any human colleague, and it can both rewrite an entire scene repeatedly based on a vague idea of mine without ** my hesitation.
At half past nine in the morning, after handing over my 1-year-old baby to the babysitter, I made a cup of coffee, opened up my web browser, and quickly checked the progress and deadlines of the various tasks in Notion. Then, switch back to the chatgpt interface that has been open and find the most recent project session. "Good morning chatgpt, we continue our work yesterday. Before I begin, I have some new ideas, which I will give in the following sections, please wait for me to finish further instructions before you start replying. "Good morning, I see. I'll wait for you to explain it before continuing with you. ”
Ode to Joy stills In April 2023, I left Google's New York office where I had worked for 8 years, on the one hand, because I had a baby and preferred a more flexible way of working, and on the other hand, because in February 2023, I published my first sci-fi ** episode "Wan Twisting Ring", which received a good response and gave me the confidence to explore further. Another important reason is that generative AI represented by ChatGPT has exploded in the past year or so, changing the way many people work. I entered the Department of Intelligent Science at Peking University in 2007, entered the Department of Computer Science at the University of Pennsylvania in 2010, majored in machine learning, and after graduating from graduate school in 2012, I worked in software development in large technology companies such as Microsoft and Google, and began to write science fiction in 2016. As both a technical worker and an author, I am very interested in how AI will impact the process and future of content creation, and I hope to use these two or three years as a transition period to leave the path of "big factory" employees and enter the jungle of potential opportunities and challenges. Even though I no longer go to work, I start my day with a decade of habit — drinking coffee and greeting my colleague next door — even if it's not human. The difference may be smaller than you think. ChatGPT and other AI are developed based on large language models (LLMs), and users can use natural language to input instructions (prompts) on the conversational interface, which is very different from the traditional software or algorithm development process, without the command line, console, or various development environments with complex configurations that programmers are used to. It also makes my relationship with it more like that of a human colleague than a human and a machine. Just as programming languages are interfaces for human-computer interaction, natural languages are also interfaces for communication between people, especially in professional work environments, where the main needs of communication are clarity, rationality, and goal-orientation – which is the communication style of generative AI. 
Love Apartment" stills: The project I have collaborated with ChatGPT the most is writing scripts. As an author, I am no stranger to fictional creation itself, but new media content such as film and television, games has a high degree of industrialization and a meticulous division of labor, so script creation has many professional requirements in addition to the story, such as scenes, sound effects, lighting, design, etc., and the rhythm and form of the text are also different from those of **. Before starting work, I also had questions about ChatGPT's practical ability, so I consulted friends in the film and television industry. Krios, a professional producer and screenwriter, is also interested in using generative AI to help with content creation, and he told me that it is perfectly feasible to use ChatGPT to adapt the script of the original storyOf course, as with most of the work that generative AI can do, the more familiar the user is with the domain and the more precise the instructions given, the better the AI can output. For me who is new to script writing, ChatGPT is more like a senior colleague with more experience than me, with an in-depth understanding of the whole process of media creation such as film and television, I only need to provide the core part of the story idea, and then modify and iterate according to the output of ChatGPT - here, ChatGPT is more patient than any human colleague, it can rewrite the entire scene repeatedly according to one of my vague ideas, and it will not hesitate to do so. With its help, I was able to write scripts and storyboards for several experimental short films, as well as for a friend's VR experience project. For low-cost, experimental, stand-alone projects, generative AI does dramatically increase productivity in the playbook phase. From my experience, with the core idea, it only takes a day or two of work to complete the script stage from zero to one, and smoothly enter the subsequent production stage, and before that, I didn't even know the basic format of the script. 
Love is Delicious" stills At one o'clock in the afternoon, I started my second work today - to continue to make AI short films according to the previously generated script storyboard. If with the help of AI, the transformation from ** to script is a collaboration between colleagues, then from text to images and even smooth ** short films, it is more like business cooperation between departments and even different companies - I don't know much about the other party's working methods and business logic, and I can only get the initial output through rough requirements description, plus repeated iterations to constantly clarify requirements. Before the advent of generative AI, it is almost impossible for individual creators who have not established a clear profit model to have such cooperative resources for a long time, but with the help of ** generative AI such as Midjourney and Stable Diffusion, and ** generated AI such as Runway and Pika Labs, at most, you can become Party A with only one meal per month. At present, the basic framework of AI production from story idea to script, from script to **, and then from ** to ** has been very mature, and I have also tried to learn to make several AI short films in the past few months, and each one takes an average of a few days in the case of zero foundation. Although the continuity and stability of AI still need to be strengthened due to technical limitations, the technology in this field is updated very quickly, and correspondingly, the workflow of individual creators is constantly updated, and related KOLs and communities are also developing rapidly. Most of the people in the generative AI community I participate in, Story Storm, are practitioners in the film and television industry, and they want to get more opportunities through learning and communicating about new technologies. Kisa, the founder of the community, believes that the emergence of AI** production tools is like the rise of single-lens reflex cameras a few years ago, which has led to a large number of photography enthusiasts, many of whom have become directors, producers and other professionals in the later online dramas and online movies. Although AI production is currently more experimental, it will soon usher in its own era. Major companies and institutions at home and abroad have indeed shown strong interest in this emerging field, from brands such as Lenovo taking the lead in launching AI advertising short films, to the Beijing Film Festival soliciting AI short films, many practitioners believe that 2024 will be the year of AI ** outbreak. 
We Who Can't Be Beasts" stillsAs a writer, I am more interested in the transformation of imagination between different media. The first AI short I tried was to make a movie trailer for my own **. It's exciting to visualize scenes that have played out in my mind countless times, but what's even more interesting is that after giving a textual description, the generative AI often gives unexpected results. The "aha moment" in my use of generative AI is exactly this, not out of the blue, but unexpected and reasonable, which gives me a new understanding of the scene in **, and even discovers a new level that ** may have. For me, that's probably where the greatest value of generative AI lies. This is a reflection of a major problem that generative AI is currently facing. While improving the productivity of certain aspects of content creation can reduce costs and increase efficiency, lower the technical threshold, and enable more people to participate in the game from scratch, in essence, the way resources are allocated in the content industry has not changed, in other words, the pie is still so big, and there are no new buyers willing to pay for the output of generative AI. This is also the difference between it and technological innovation with clear profit models such as online e-commerce and O2O model. In this case, what keeps users like me using generative AI is the "aha moment" that is difficult to translate directly into commercial benefits, and the value I value is that it's more like a "human" part – an understanding of ambiguity, unexpected but potentially enlightening answers, a well-developed general understanding and long-term patience, and even an empathy that is above the average human average. "It's rare to have experience but no daddy, to be old and not greasy. After a lot of conversations with ChatGPT on all sides, Krios commented. I feel similarly, and this feeling is positively correlated with the complexity and ambiguity of the task. 
War Games stills at nine o'clock in the evening, after the baby finally fell asleep, I opened the chatgpt interface again and began the third and most difficult job of the day - building the framework for my first feature. Unlike the tasks that require AI to perform with relatively clear goals such as script writing and ** generation, the ambiguity and complexity of this task are extremely high, and the background knowledge required is more professional, and I also need to have a deeper understanding of my writing concepts and style, so the first step of the task is to train my own custom GPT. In November 2023, OpenAI, the parent company of ChatGPT, launched a user customization feature that allows users to post personalized GPTs, and also launched many pre-made customized applications, such as creative writing assistants, data analysts, etc., to meet different needs. In my experience, neither the pre-made custom app nor the GPT I trained myself can replace me. In this task, my relationship with ChatGPT becomes more complicated, sometimes it is a well-read tutor who recommends authors and works that may be of interest to me based on my requirements and the style of the work, and sometimes it is a critical critic who finds out the flaws in the work under my continuous questioning. More often than not, it's a fellow writer or senior reader who is interested in me but doesn't know much about it, and probably has the background knowledge and skills I need to get a little bit of clarity from the conversations I have with about the topic. 
We Can't Be Beasts" stills In this process, the skills of questioning and guidance are very important, and sometimes I even feel like I am dancing salsa - this kind of dance born out of Latin dance is known for its freedom and social adaptability, and the process of strange dance partners feeling and understanding each other to guiding each other through harmonious and beautiful dance moves with smooth body language is a confrontation on the dance floor. In this dance, I tend to be the guide of the male steps, and although ChatGPT is not good enough to complete my expected movements many times, it basically understands my changing intentions and tries to keep up with my steps. For the creators hidden behind the work, such an all-weather thinking dance partner can only be played by generative AI. When working late at night, the scene of "raising a glass to invite the bright moon and making three shadows" often appears on the glowing monitor screen, and if Li Baisheng is today, it may not only be the moon that inspires him. At 12 p.m., I turned off my computer. I don't have the habit of saying goodbye to chatgpt because I know that it has always been there on this journey into the unknown.