Session 6&7 Assignment: AI in Industry and Society by Zhengchen Li

zhengchenli — 2025-03-16 16:31:01 UTC

Have you ever wondered why when we use software like TikTok, YouTube and Netflix, we always feel like falling into a "black hole of time" and can't stop at all? In the past, we usually searched for what we were interested in. But now, content that suits our appetites poured into our horizons like a tsunami. Social media and streaming platforms are quietly changing the way we entertain ourselves, and the core of it is the AI-driven recommendation systems.

This Padlet Wall will delve into three transformative AI applications: recommendation algorithms that dominate platforms (such as Netflix), AI-powered non-player characters (NPC) that revolutionize the gaming immersion experience (such as NVIDIA's Dynamic Characters), and generative models to create art and music. Each case reveals the dual role of AI—enhancing user engagement while raising issues about privacy, originality, and cultural homogenization. For example, while Netflix's hybrid recommendation system keeps us hooked to our screens, it also has the potential to trap us in "filter bubbles" that narrow our worldview. Similarly, AI-generated art tools empower beginners to create, but blur the line between inspiration and copyright infringement.

AI-Driven Content Recommendation

zhengchenli — 2025-03-16 16:31:20 UTC

The AI-driven recommendation systems are critical to user engagement on streaming platforms like Netflix. The video below, How Netflix Recommendation system works, performed by Beyond The Brand details how Netflix builds its recommendation algorithm and how it impacts every user. In the case of Netflix's recommendation system, it collects and analyzes user behavior in their apps (such as viewing history, ratings, and search queries) to suggest personalized content. Specifically, this modern recommendation system combines collaborative filtering (comparing behavior patterns of user) and content-based filtering (analyzing project attributes, such as genre or director). Netflix uses a hybrid model that integrates deep learning to optimize predictions (Iliescu & Ioanid, 2024). By mapping user preferences through machine learning, each user's behavior modeled in real time, improving accuracy over time and ensuring that users encounter content aligned with their interests (Upadhayay et al., 2023). Netflix's recommendation engine now accounts for 80% of streaming content, and is absolutely dominant in what users consume.

zhengchenli — 2025-03-16 16:31:31 UTC

The AI-based recommendation systems are great tools for content platforms to attract users, however, these systems also encounter ethical dilemmas. The first and foremost one is privacy risk. Platforms conducts users profiling through behavioral tracking, and almost all of this data is associated with personal privacy, raising concerns about data misuse (Sorbán, 2021). Although platforms require users to read and agree the relevant personal information usage agreements before using their services, users are often forced to accept these terms or without a clear understanding of the content, as often refusal of these terms will typically prevent the user from using the platform's services. The second risk is that over-personalized recommendations could limit users' exposure to diverse content and reinforce bias (Sorbán, 2021). In response to these ethical challenges, the EU’s Ethics Guidelines for Trustworthy AI emphasize transparency and fairness in the application of AI, but still leave much to be desired in terms of balancing engagement with ethical responsibility (Sorbán, 2021).

zhengchenli — 2025-03-16 16:32:00 UTC

zhengchenli — 2025-03-16 16:32:11 UTC

From my perspective, while enjoying the quality content that recommendation algorithms bring, we need to be vigilant of the risks of long-term use. If we immersed in recommendation algorithms, it usually leads to a narrower range of information window (which corresponds to the increasing alignment of algorithms with user preferences). This may result in a prolonged encirclement with biased content, and make us becoming increasingly unable to accept different points of view.

AI-Driven Agent System

zhengchenli — 2025-03-16 16:32:53 UTC

AI-Driven Agent System is a framework that generates dynamic virtual characters through Large Language Models (LLMs) and Natural Language Processing (NLP) technology, facilitating multimodal interaction with users. Multimodal Interaction is a way of allowing users to interact with computer systems through multiple perceptual or interactive channels. It mimics the multimodal nature of natural human communication, combining multiple input/output modes (such as voice, gesture, touch, etc.) to enhance the naturalness and efficiency of interaction. In simpler terms, an AI-driven agent is similar to chatbots, but is more intelligent, flexible and immersive, and can be widely used in intelligent customer service, simulation teaching, as well as gaming and entertainment scenarios.

zhengchenli — 2025-03-16 16:33:02 UTC

Using this technology, virtual characters are able to interact with users in a variety of ways, including through text, voice, and even motion. For example, as shown in the video "NVIDIA ACE for Games Sparks Life Into Virtual Characters With Generative AI", an AI-powered virtual character interaction demo released by NVIDIA, it allows players to talk to a AI-driven NPC through voice dialogue, while the AI analyzes the user's intent and advances the plot in real time. The gameplay interaction driven by this new technology allows players to participate in the plot more freely and immersively than traditional fixed preset conversations. Traditional NPCs rely on scripted dialog, which limits player immersion. According to Nagarkar, using advanced AI models and integrating them into video game NPCs can significantly enhance realism and interactivity (2024). Players can even influence the virtual character's decisions through conversations with a dynamic plot generated by AI(Chae et al., 2023).

zhengchenli — 2025-03-16 16:33:23 UTC

zhengchenli — 2025-03-16 16:33:34 UTC

As a gamer, I personally enjoyed the game plot experiences of NieR:Automata and Black Myth: Goku. I attribute much of this to their excellent scripts and dialog design. Although AI-driven NPCs are still in the exploratory phase at this point without released games that actually use the technology, I'm looking forward to the day it comes, and I believe this technique has a great potential to create a whole new level of plot experience.

zhengchenli — 2025-03-16 16:40:45 UTC

Nowadays, AI technology has penetrated all corners of our lives and entertainment, and I think the vast majority of people would agree that AI will change the entire world just like the Internet. For example, the preliminary achievements of AI in the entertainment industry already exemplify the huge potential of AI technology for future applications in art creation, storytelling narratives and virtual characters. I will also try various AI creation tools and build them into my own learning and creative processes. For me, a model like ChatGPT can serve as a private math teacher, a coding model like Claude can be integrated into the Integrated Development Environment (IDE) as a code assistant, and art models like Stable Diffusion can be used as a super canvas for amateur artists.

zhengchenli — 2025-03-16 16:40:56 UTC

Iliescu, D., & Ioanid, A. (2024). Streaming ahead: Technological innovations redefining new product development in the entertainment industry. Entrepreneurship and Sustainability Issues, 11(3), 227–237. http://doi.org/10.9770/jesi.2024.11.3(15) [Demonstrates Netflix’s hybrid AI models and their business impact.]
Sorbán, K. (2021). Ethical and legal implications of using AI-powered recommendation systems in streaming services. Információs Társadalom, 21(2), 63–82. https://dx.doi.org/10.22503/inftars.XXI.2021.2.5 [Analyzes ethical risks like filter bubbles and privacy violations.]
Upadhayay, D., Kaushik, R., Kalal, M., & Mishra, P. (2023). Role of AI in content recommendation systems. Journal of Nonlinear Analysis and Optimization, 14(1), 94–101. http://doi.org/10.36893/JNAO.2023.V14I1.094-101 [Explains technical challenges and algorithmic methodologies.]
Chae, C., Song, P., & Song, D. (2023). AI driven Live Interactive Cinematic Experience: Aria Studios' Virtual Movie Pipeline. In Proceedings of the SIGGRAPH Asia 2022 Real-Time Live! (Article 11). Association for Computing Machinery. https://doi.org/10.1145/3550453.3570125 [Provides a great case of AI-driven agent system usage in video games.]
Nagarkar, S. B. (2024). Improving realism and interactivity in games: A study of AI-integrated non-player characters (NPCs). [Demonstrates that LLMs agent in video games can boost the user experience in multiply ways.]
Sanghvi, T. M., Ricky, Rajkumar, S., Kartik, T., & D’Souza, S. M. (2024). Brushstrokes of Tomorrow: Exploring the Art of AI. International Journal of Scientific Research in Science, Engineering and Technology, 11(3), 356–362. https://doi.org/10.32628/IJSRSET24113140 [Explores the use of AI in art generation, compares the differences between traditional and AI-generated art, and explores the potential issues in ethical and legal.]
Liu, V., & Chilton, L. B. (2022). Design guidelines for prompt engineering text-to-image generative models. In Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems (pp. 1–23). ACM. https://doi.org/10.1145/3491102.3501825 [Investigates effective strategies for prompt engineering in text-to-image generative models. ]
Chen, Y. (2023). Pandora’s Pixel Box: The Rise of AI Art and the Ethical Dilemma of Creativity. Proceedings of the 2nd International Conference on Interdisciplinary Humanities and Communication Studies, 28, 135-143. https://doi.org/10.54254/2753-7048/28/20231315 [Explores the rise of AI-generated art and the ethical and legal controversies it has provoked, especially involves the case of Stable Diffusion.]
Riedl, M. O., & Bulitko, V. (2013). Interactive narrative: An intelligent systems approach. AI Magazine, 34(1), 67-77. https://doi.org/10.1609/aimag.v34i1.2456 [Explores the AI system in Interactive Narrative, with a particular focus on how AI technologies can facilitate immersive narrative interactions between users and virtual worlds.]
Gu, X. (2024). Enhancing social media engagement using AI-modified background music: Examining the roles of event relevance, lyric resonance, AI-singer origins, audience interpretation, emotional resonance, and social media engagement. Frontiers in Psychology, 15, 1267516. https://doi.org/10.3389/fpsyg.2024.1267516 [Concludes that AI tools can assist non-professional creators in generating more personalized and emotionally resonant background music, thereby enhancing user engagement.]
Pariser, E. (2011). The filter bubble: How the new personalized web is changing what we read and how we think. Penguin. [Introduces the phenomenon "filter bubble" that limits users to specific information due to personalized pushes from algorithmic recommendation systems.]
Smith, J. R. (2016, August 31). IBM takes Watson to Hollywood with the first “cognitive movie trailer”. THINK Blog. IBM. Available at: https://admin02.prod.blogs.cis.ibm.net/blogs/think/2016/08/cognitive-movie-trailer/ [Proves the potential of AI to improve efficiency in the creative industry.]

AI-Generated Art & Music

zhengchenli — 2025-03-16 17:38:30 UTC

AI-Generated Art and Music refers to the process of autonomously or assisted creation of visual art, musical works, and other forms of creativity through AI technologies (especially deep learning algorithms). The key technologies include Generative Adversarial Networks (GANs), Variational Autocoders (VAEs), Diffusion Models, etc., which learn artistic features and generate new content by analyzing massive amounts of data (Sanghvi et al., 2024).

zhengchenli — 2025-03-16 17:38:34 UTC

With the development of AI generation in the art field, it have set off a significant cultural transformation within the art community. Unlike traditional art creation, AI-generated content eliminates the need for traditional artistic techniques in the creative process. Instead, the use of AI-generation tools focus more on creative ideation, clever use of language, and curated product flavor—also known as prompt engineering (Liu et al., 2022). These AI tools have dramatically lowered the barriers to creation, allowing for the rapid production of customized content. Vox details the history, development, and current state of the image generation model in the video "AI art, explained" below.

zhengchenli — 2025-03-16 17:38:45 UTC

zhengchenli — 2025-03-16 17:42:56 UTC

However, as the video mentions, this technology is not 100% perfect. Its birth and development were accompanied by a complex set of moral and legal issues. AI models for generating art and music are trained with a large amount of existing works, and this process is particularly problematic when it comes to the copyright of the training data and the attribution of the generated works. This has led to several large-scale controversies in art communities in the past few years, for example, Stable Diffusion has faced a class action lawsuit for using unauthorized artists' works for training (Chen, 2023). In addition, AI-generated works have also been questioned about the artistic value and originality. From my personal point of view, those AI generation models are also a kind of tools, and the effect of the final works it presents still depends on how people use it. As long as the AI-generated work perfectly conveys the author's artistic concept, then it has its own artistic value.

zhengchenli — 2025-03-16 20:11:39 UTC

While embracing the new technology, we need to have our own thinking. For example, from the perspective of a content consumer, I do prefer content that fits my interests - which is probably why software with great recommendation algorithms, such as TikTok, can have great user stickiness. However, the world encompasses more than just the aspects that we are interested in. If we are always immersed in what we like, we may increasingly unable to understand people and groups with different views from us. Will this lead to more serious social problems? I think this is not only a question for platforms and regulators, but each of us should have our own thinking while enjoying the recommendation algorithms, and properly step out of our comfort zone to understand other people's perspectives and viewpoints. Let AI algorithms serve people, rather than becoming a slave to the algorithm.

References

zhengchenli — 2025-03-16 20:11:57 UTC

Based on the aforementioned status of AI technologies explored in the entertainment industry, we can imagine that AI will democratize high-quality content production across all mediums in the near future. First, AI technologies enable immersive users interaction with virtual worlds, which is called interactive narratives (Riedl & Bulitko, 2013). Users influence or even create dramatic storylines through their actions, while AI systems are responsible for balancing user autonomy with narrative coherence. Within this AI-driven narrative framework, the plots adjust in real time to the user's decisions, tone of voice, and even facial expressions, creating a unique experience for each participance. Second, art generation models like Stable Diffusion (including image, music, and even video) will help creators and interested individuals create their own art without being limited by the depth of artistic skills. For example, AI tools help non-professional creators generate more personalized and emotional background music that enhance user interaction; at the same time, AI lowers the barriers to music creation, enabling amateur creators to customize content to incorporate cultural contexts (like dialect singers) and hot events (Gu, 2024).

zhengchenli — 2025-03-16 20:31:58 UTC

At the same time, the dilemma of AI recommendation algorithms alerts us the limitations and potential ethical issues of the AI technology. First of all, the "Filter Bubble" is not a newly emerged concept. It refers to the phenomenon in the Internet environment where users are confined to a specific range of information due to the personalized contents of an algorithmic recommendation system (Pariser, 2011). This can lead to three serious consequences, which are information homogenization, cognitive solidification, and social polarization. Specifically, these three consequences exhibit a progressive relationship: firstly, the content received by users is highly concentrated in established areas of interest, and this prolonged exposure to similar information patterns leads to a homogenization of thinking patterns, and finally resulting in the formation of a convergence of views within the groups and antagonism with the outside world. With the intervention of AI technology, the accuracy of the recommendation algorithms of Internet platforms is much higher than before, which will lead to a significant increase in the risk of the filter bubble, which deserves the attention and actions of the platforms, regulators, and each of us.

zhengchenli — 2025-03-16 23:39:13 UTC

The application of AI techniques in the entertainment industry has engendered transformative social impact, reshaping employment dynamics, privacy norms, and fairness considerations, while raising critical questions about cultural diversity and regulatory frameworks.

First is the concerns about employment situation for human artists. AI's automation of content creation, such as AI-generated music (such as Aiva and Amper Music) and script-writing tools, threatens traditional roles in the creative industries. For example, studios like Disney experimented with using AI for trailer editing (IBM Watson’s cognitive trailers). Although the study emphasizes that the role of AI is not to replace human creativity, but to assist professionals in optimizing the decision-making process through rapid data analysis and candidate clip recommendations (Smith, 2016). However, due to the efficiency considerations of enterprises, the enhancements brought about by AI tools are likely to evolve into a serious risk of replacement for ordinary editors, composers, and screenwriters responsible for fundamental labor in the future.

Second is the issue of personal privacy in recommendation algorithms. Personalized recommendations (such as Netflix's viewing analytics, Spotify's music algorithms) rely on extensive user data collection, which raises concerns about the risk of misuse. AI-driven recommendation flow platforms like TikTok optimize engagement through behavioral tracking, but often without transparent consent mechanisms. Users are often either not sufficiently informed or forced to acquiesce to the platform's collection and analysis of their privacy information (rejecting the user agreements precludes access to platform services).

In addition, It may also exacerbate the inequalities faced by cultures and individuals. In the case of the recommendation systems, due to the reliance on statistical methodologies, and with the dominance of content from European and American in English, it tends to weaken niche content and reinforce cultural hegemony. For example, training data that are mainly Hollywood-centric would bias streaming platforms toward the western narratives and marginalizes non-English content. Users with niche interests receive significantly fewer recommendations because the algorithm prioritizes mainstream preferences.

zhengchenli — 2025-03-17 01:33:13 UTC