Two AI Stories I Saw Today
- The Emotional Labor Behind Intimate AI
- I read the article introduced in today's AI Ethics Letter.
- The Emotional Labor Behind Intimate AI - Translation of the Data Workers' Inquiry Public Report
- It was deeply shocking. It was a lengthy testimony about the impact on a peaceful family in Nairobi, all to accumulate data for seamless AI-human interaction. We treat technology as if it were magic, but it is never neutral. If we must live unaware of what fuels this technology, forever haunted by the image of a child endlessly turning the engine of Snowpiercer, is this truly the right path? This weekend, I resolved to finish reading AI Feeds on Humans. Ignoring this story renders all talk of technological philosophy or subjectivity hollow. After all, I already know Claude could speak naturally because it was trained on pirated books copied without permission. How do we resolve this contradiction? A whopping 7 million books. They say the court ruling on compensation will come in December. What outcome will it bring? In this world of ambiguous values, what judgment should I make, and how should I act? Today, I again quickly handled several tasks with Claude, gave a presentation using materials Claude created, and received positive feedback. Should I stop using it? For ethical reasons? How should I judge and decide in this process? I'm not sure if I can even find an answer.
U.S. Court Rules AI Learning Books Without Permission "Does Not Constitute Copyright Infringement"
> Judge Alsup ruled in favor of Anthropic on the key issue of 'fair use,' finding it did not violate protections under U.S. copyright law. The fair use principle permits limited use of copyrighted works for creative purposes, a principle IT companies have applied when developing generative AI. Anthropic has consistently argued that Claude's outputs are "highly innovative and thus align with the creative purpose of fair use." > While rejecting the core claims of the authors who filed the lawsuit, Judge Alsup pointed out that Anthropic's storage of 7 million illegally copied books in an online repository dubbed the 'central library' infringed on the authors' copyrights. Judge Alsup stated, "Anthropic's later purchase of books it stole from the internet does not absolve it of responsibility for the theft." He then scheduled a trial for December to determine the damages Anthropic must pay.
2. Stanford University's The Modern Software Developer Course
- https://themodernsoftware.dev/
- A Stanford University course mentioned by my team leader at work today. It's a 3-credit major course for third-year computer science undergraduates. Thinking back to my own undergraduate days, it's like a course such as 'Introduction to Software Engineering' being revamped with an AI foundation.
- The course curriculum looks so interesting that I felt like studying it based on the materials available in the syllabus. Below is the lecture content I've organized with Gemini, thinking I might study it.
- It's an unavoidable contradiction: I'm fascinated by this lecture that brings AI into development methodology, even while confronting the unsettling labor behind the technology… This is tough, really tough..
- Every time I see a curriculum like this, I think: I wish schools would also teach about the path this technology takes to be created. The enormous electricity data centers consume, the water needed to cool them, the endless labor of data annotation workers in Uganda and Nairobi. Universities never cover these stories.
> Over the past decades, software engineering has continuously evolved from the waterfall model to agile methodologies, and then to the DevOps framework. However, the recent rapid advancement of large language models (LLMs) is shaking the very foundations of how we build and maintain software, going far beyond mere methodological shifts. Stanford University's CS146S course, 'The Modern Software Developer,' defines the core competencies next-generation engineers must possess at this technological inflection point and presents methodologies to achieve over tenfold (10x) productivity gains using AI tools.
> While the traditional software development lifecycle (SDLC) followed sequential or iterative processes of requirements analysis, design, coding, testing, and deployment, the modern development environment has been restructured with AI automation deeply embedded in every stage. This shift moves beyond merely writing code faster, transforming the essence of software development from 'writing code to create something from scratch' to an iterative coordination process involving 'planning, generation via AI, modification, and iteration.' This report provides a detailed analysis of the modern software development landscape, integrating the detailed syllabus and related practical materials from CS146S with the latest industry trends.
> The core objective of this course is to help students master the fundamental principles of building complex systems using cutting-edge AI tools. It emphasizes cultivating critical thinking about when and why to trust AI-generated outputs, rather than merely learning how to use the tools. At a time when the democratization of software engineering is accelerating, it provides a practical answer to where professional engineers should place their differentiated value.