What I've Been Reading (9) (2023-04-18)
Software Engineering
Mainframe → Relational Databases → Data Warehousing / Analytics → Collaborative Databases. I’d argue that this is an increase of options, not an evolution of the best known way to handle data. Sometimes having everything in one place is valuable. Next year, SQL turns 50 years old!
Principles useful when doing UX work. Nothing too crazy, but fun to click through.
Fun Tech
https://tonyzhaozh.github.io/aloha/
Little robot hands can put on contacts, open containers, and put on shoes. Eventually we’ll get to the world where I never need to leave my bed.
https://github.com/ange-yaghi/engine-sim
Accurate engine simulation designed to make noise for video games. The youtube video is worth a watch.
Science
https://www.experimental-history.com/p/science-is-a-strong-link-problem
Weak-link problems: most value comes from minimizing failures (food quality, computer security, nuclear access). Strong-link problems: most value comes from the best thing being better (music, friendships, venture capital). Science is a strong link problem: It’s more important to have some great research than to avoid any bad research.
Security
https://kentindell.github.io/2023/04/03/can-injection/
This made the rounds a few weeks ago - if you were like me and got scared off by the length, it’s worth reading - it’s not too technical as far as posts like these go. Since internal CAN messages are unauthenticated, hackers can inject their own messages and start your car right on up. Only costs $10 or so in parts.
AI
https://www.izzy.co/blogs/robo-boys.html
Just a fun project. AI-powered group chat that knows all about you and the boys. Is when LLMs understand inside jokes when it starts getting creepy? Nah, just funny.
https://jacobbuckman.substack.com/p/deep-learning-is-better-than-linear
Lots of graphs to compare deep learning and linear regression. Did I mention there are lots of graphs?
https://magazine.sebastianraschka.com/p/understanding-large-language-models
Links to papers to help understand the history/math/tech of language models. Starting from 2014.