What FYSK is Reading
Some articles and books recommended by our community to kick off your new year
Have you ever examined an alien’s brain? Are you playing infinite games? How do you translate your technical prowess into power?
The folks who make up the FYSK community are building tomorrow's AI applications. So we asked them what is inspiring their curiosity.
“What’s one piece you read this past year that changed your perspective and why?”
Grab a cup of coffee, throw a log on the fire and read along with us!
John Bohannon, Senior Director of Data Science at Primer.ai
Sparks of Artificial General Intelligence: Early experiments with GPT-4
This paper gave me shivers. It's a psychological examination of an alien mind. It asks questions that we'd never have dreamt to ask just 2 years ago. I've generally been an AGI skeptic, especially when it comes to transformer language models. I remember using GPT-3 (before instruction tuning) and thinking, “well there's your proof that more-data-bigger-transformer won't get us to AGI”. And then OpenAI changed my mind. I remember this paper feeling like an inflection point. Like maybe this is actually happening. Not just maybe in my lifetime. Maybe SOON.
Dennis Xu, Founder at Mem
Software Localism, or the Law of Two Peters
There’s just so much software in the world that is yearning to be built. We will probably see 4-5 orders of magnitude more software applications that people rely on daily by 2030!
Swetha Mandava, Founder at Truco, ex-NVIDIA Senior Deep Learning Engineer
Finite and Infinite Games
This year, I read about this concept from James P. Carse's book. It provides a great framework for ideas we've probably always known. In this model, finite games have clear rules and the objective is to win, while infinite games have no fixed rules, aiming instead to continue playing and evolving. I often find myself thinking about whether something I'm doing is a finite game or an infinite game. In fact, most things seem like infinite games to me now: building a company or even an aggressive game of Catan (if you want to continue playing with family). 🙃
Nyla Worker, Director of Product Management at Convai, ex-NVIDIA Product Manager
The 48 Laws of Power
A couple of my peers who have really shone in the corporate world recommended this book to me. The book itself is morally dubious and outdated in how power is attained today. However, the book has enabled me to pattern-match different power games in which we all participate knowingly or unknowingly. In a world where we have AIs that can converse in a natural way, I do wonder at what point AIs will be usable to extend power plays and manipulative behaviors.
Emma Qian, Founder at Stealth AI Startup
Textbooks Are All You Need
This paper surprised me with how well a small model performs when it's trained on high quality data. I became more convinced that small, specialized models will be more important in the future. I also found it interesting to see how the paper's approach used large models to create high-quality synthetic training data.
We’d love to hear from you. What are you reading this year? What’s in your queue to read “coming up”?