Issue #278

April 16, 2025

Dear Reader,

Welcome to the Apr 16th issue of the Data Science Briefing!

The latest blog post on the Epidemiology series is now out: Demographic Processes. In this post we explore how to include birth and death rates in your epidemik models. Check it out!

This week we cover how OpenAI is reportedly developing a new social network centered on ChatGPT’s image generation capabilities, aiming to compete directly with platforms like Elon Musk’s X and Meta’s Facebook and Instagram. This initiative, still in prototype form, could allow OpenAI to harness real-time social data to train its AI models further. Meanwhile, Google has released a comprehensive 69-page whitepaper on prompt engineering, offering best practices to optimize interactions with generative AI models—a valuable resource for data scientists and ML practitioners seeking to refine model outputs. On the cybersecurity front, GitHub suffered a significant supply chain attack compromising CI/CD secrets across over 23,000 repositories, exposing sensitive credentials such as AWS keys and personal access tokens, highlighting the growing risks in software supply chains and the critical need for vigilant security monitoring. These developments underscore the dynamic intersection of AI innovation and security challenges shaping the current technology landscape.

On the academic front, this issue showcases a dynamic interplay between emergent hardware innovations and advanced analytical techniques. Researchers are breaking new ground with self-organizing neuromorphic nanowire networks that operate as stochastic dynamical systems, heralding a new era in brain-inspired computing. At the same time, fresh perspectives on complex systems emerge from studies of strange attractors, deepening our understanding of how unpredictable patterns arise in interconnected networks. For those looking to refine their analytical toolkit, a practical guide offers ten quick tips for leveraging Bayesian statistics, making sophisticated inference more accessible. Complementing these technical advances, research into political ideology and trust in scientists underscores the societal dimensions that critically shape the reception and impact of scientific discovery. Lastly, innovative strides in fine-tuning language models with collaborative and semantic experts illustrate the seamless convergence of theory and application, driving the future of intelligent systems.

This week’s book is “Prompt Engineering for LLMs” by J. Berryman and A. Ziegler. You can find all the previous book recommendations on our website. In this week’s video, we have a tutorial on Instruction Finetuning.

Data shows that the best way for a newsletter to grow is by word of mouth, so if you think one of your friends or colleagues would enjoy this newsletter, go ahead and forward this email to them. This will help us spread the word!

Semper discentes,

The D4S Team

"Prompt Engineering for LLMs" by J. Berryman and A. Ziegler is an essential resource for anyone working with large language models. The authors expertly position prompt engineering not merely as writing effective prompts but as a crucial component throughout the entire application development lifecycle. By balancing technical depth with practical accessibility, they create a guide that serves both newcomers and experienced practitioners in the rapidly evolving AI landscape.

The book's greatest strength lies in its practical techniques, which go beyond basic prompt crafting. Readers will discover innovative approaches, such as using log probabilities to quantitatively assess completion quality, generating multiple outputs at varying temperatures, and structuring prompts with multiple roles to enhance focus and relevance. Particularly valuable is the "Little Red Riding Hood Principle," which emphasizes aligning prompts with a model's training patterns to achieve optimal responses.

Beyond techniques, Berryman and Ziegler offer crucial insights into real-world application strategies, including how teams like GitHub Copilot incorporate user feedback for continuous improvement. The authors skillfully explain complex concepts like tokenization and auto-regressive generation while maintaining accessibility for developers who might otherwise struggle with the non-human communication style of LLMs. This balanced approach makes the book an indispensable guide for anyone aiming to build robust, efficient LLM-powered applications in today's AI-driven technological environment.

Prompt Engineering for LLMs

1. OpenAI is building a social network [theverge.com]
2. Google Prompt Engineering Guide [kaggle.com]
3. Google Is Winning on Every AI Front [thealgorithmicbridge.com]
4. What is Entropy? [jasonfantl.com]
5. These are the most-cited research papers of all time [nature.com]
6. GitHub suffers a cascading supply chain attack compromising CI/CD secrets [infoworld.com]
7. Cross-entropy and KL divergence [eli.thegreenplace.net]
8. Controlling Language and Diffusion Models by Transporting Activations [machinelearning.apple.com]

• Self-organizing neuromorphic nanowire networks as stochastic dynamical systems (G. Milano, F. Michieletti, D. Pilati, C. Ricciardi, E. Miranda)
• Political ideology and trust in scientists in the USA (V. Gligorić, G. A. van Kleef, B. T. Rutjens)
• Strange attractors in complex networks (P. Villegas)
• Ten quick tips to get you started with Bayesian statistics (O. Gimenez, A. Royle, M. Kéry, C. R. Nater)
• Estimating Re and overdispersion in secondary cases from the size of identical sequence clusters of SARS-CoV-2 (E. B. Hodcroft, M. S. Wohlfender, R. A. Neher, J. Riou, C. L. Althaus)
• A generalized higher-order correlation analysis framework for multi-omics network inference (W. Liu, K. A. Pratte, P. J. Castaldi, C. Hersh, R. P. Bowler, F. Banaei-Kashani, K. J. Kechris)
• Fine-Tuning Language Models with Collaborative and Semantic Experts (J. Yang, B. Hui, M. Yang, L. Zhang, Q. Qu, J. Lin)

Build an LLM from Scratch 7: Instruction Finetuning

April 16, 2025

​