I apologize for this being a day late. I’ve overhauled a bit to make it more easily readable. Enjoy! ?
Here's a comprehensive AI reading list from this past week. Thanks to all the incredible authors for creating these helpful articles and learning resources.
I put one of these together each week. If reading about AI updates and topics is something you enjoy, make sure to subscribe.
Society's Backend is reader supported. You can support my work (these reading lists and standalone articles) for 80% off for the first year (just $1/mo). You'll also get the extended reading list each week.
A huge thanks to all supporters.
What Happened Last Week
Here are some excellent overviews of what happened in AI this past week:
AI Agents Weekly by to learn more about how AI agents are being used and developments in that space.
’s AI Roundup. This is an article I read each week.
The Weekly Kaitchup by . This is a new one I’m adding that this week. It contains succinct overviews of important AI topics.
’s AI Roundup is an overview of AI papers and tools from the past week.
The Batch is an excellent overview of AI developments and their impact from Andrew Ng each week. I highly recommend it.
Papers Podcast
ML papers are difficult to keep up with. Here’s this week’s NotebookLM-generated podcast going over important papers you should know:
Last Week's Reading List
Reading List
Evaluating feature steering: A case study in mitigating social biases
The study explores feature steering as a method to reduce social biases in AI models while examining its impact on overall model performance. Researchers found a "sweet spot" for steering that can decrease biases without significantly harming capabilities, but also noted unpredictable effects in model outputs. Further research is needed to understand the effectiveness and limitations of feature steering across various contexts.
A Technical Perspective: Has Google Search Gotten Worse?
The article discusses complaints about the quality of Google Search and Search Ads, emphasizing the sophisticated machine learning systems behind them. It explains that while many users feel search results have declined, Google constantly adjusts its algorithms to improve user experience and reduce SEO manipulation. The author believes that when done correctly, Search Ads can provide a beneficial second search feed without degrading user experience.
How NotebookLM Was Made
By
Raiza discusses the development of NotebookLM, highlighting the unexpected enjoyment from user audio overviews. They mention the need to simplify features that weren't widely used and the importance of creating engaging content. The team is still experimenting and learning how users interact with the tool, focusing on continuous improvement.
Unbounded: A Generative Infinite Game of Character Life Simulation
"Unbounded" is a generative infinite game where players create and interact with a custom wizard character named Archibus. The game adapts in real-time, allowing for spontaneous storytelling and a variety of actions in different environments. It uses advanced AI models to enhance character consistency and narrative coherence, offering a unique life simulation experience.
Why everyone should learn to code
By
Learning to code is a valuable skill that can greatly enhance problem-solving abilities and lead to high-paying job opportunities. It teaches you to think abstractly, breaking down complex issues into manageable parts. However, mastering coding requires practice and a shift in thinking, as many resources focus only on writing code rather than understanding the underlying concepts.
Security research on Private Cloud Compute
Apple's Private Cloud Compute (PCC) enhances privacy and security for AI processing by allowing researchers to verify its claims. They have made tools and source code available for independent analysis and expanded their security bounty program to reward findings related to PCC vulnerabilities. This initiative aims to build trust and improve PCC's security and privacy features with community involvement.
SynthID
SynthID is a tool developed by Google DeepMind that embeds invisible digital watermarks in AI-generated content like images, audio, text, and video. This technology helps users identify AI-generated material, promoting trust and safety in information. Currently in beta, SynthID is being integrated into various products to enhance responsible use of AI.
How Jacob Collier helped shape the new MusicFX DJ
Jacob Collier collaborated with Google to create MusicFX DJ, a tool that helps anyone make music using AI. This tool allows users to mix different instruments, genres, and emotions to create a continuous flow of music. Users can also share their music sessions with others, encouraging collaboration and creativity.
State of the software engineering job market in 2024
By
The software engineering job market in 2024 shows signs of growth after a decline in 2023, with many companies actively hiring. Amazon, Oracle, and TikTok are among the top recruiters, while other big tech firms like Meta and Google have reduced their workforce. New trends indicate a shift in hiring sources, with companies like OpenAI and Anthropic looking to attract talent from major tech giants.
