As the summer holiday season winds down, let's turn our attention to the hot topic of Generative AI and Large Language Models (LLMs). There is immense potential and exciting opportunities in this field, and I am curious to see how it will shape our world. In this month's Dev Breakfast you’ll find a practical guide to building successful products around LLMs, a walkthrough of the end-to-end engineering process of an LLM-based application, and articles to deepen your understanding of the core concepts. Join us as we explore these cutting-edge technologies.
Bonus - scroll down to get the Futurice's Generative AI working paper.
Meet our curator of the month
Larisa is a Software Developer at Futurice with a solid seven years of experience in creating modern software. She enjoys adapting technologies to real-world problems and learning new abstractions to find better solutions. In her spare time, she dives into dog behavior and psychology, which she uses to train her two-year-old Labrador Retriever.
By Eugene Yan, Bryan Bischof, Charles Frye, Hamel Husain, Jason Liu, Shreya Shankar
A group of authors from a variety of backgrounds created a comprehensive practical guide to building successful products around LLMs, drawing from their own experiences and pointing to examples from around the industry. They cover tactical, operational, and strategic perspectives. I believe everyone curious about LLMs will find something interesting and useful in this guide.
By Aeree Cho, Grace C. Kim, Alexander Karpekov, Alec Helbling, Zijie J. Wang, Seongmin Lee, Benjamin Hoover, Duen Horng Chau
All cutting-edge Generative AI models, including both language and image models, are built on the transformer architecture. For those interested in a deeper understanding of this technology, the Transformer Explainer is an excellent open-source interactive visualisation tool. It uses the GPT-2 model as an example to help users understand complex Transformer concepts by enabling smooth transitions across abstraction levels of mathematical operations and model structures. Start your journey with this introductory video. Additionally, I recommend another visual walkthrough that effectively demystifies Transformers.
Security vulnerabilities of LLM based applications are a critical and intriguing subject. The concept of "prompt injection" is one aspect that has particularly drawn my interest. In this blog post Simon Willison, who actually coined the term, provides a concise summary of the problem, complete with examples. While newer models are becoming more resilient to this class of vulnerability, they are not yet completely immune.
The concept of embeddings is also fundamental to understanding LLMs. Embeddings are essentially vector representations of text that encapsulate its meaning, allowing us to measure semantic similarity between different pieces of text. This feature has a wide range of practical uses.
For a real-world example of a project centred around embeddings, take a look at Wilson Lin's blog post. He details his process from starting with zero data and no code to developing interactive tools for search, analysis, and spatial visualisation. His journey involves working with millions of Hacker News comments, tackling various challenges, and discovering solutions along the way. This case study sheds light on the power and applicability of embeddings in many areas.
Download Futurice's dynamic, evolving resource - an educationalworking paper created to keep pace with the rapid advancements of Generative AI.
Know your curator
How do you stay updated on the latest industry trends and technologies?
To stay updated with the latest tech news, I use a mix of sources. I browse trending topics on Hacker News, delve into conversations across various Reddit channels, and follow a selection of reliable tech personalities on Twitter. Despite this, the most insightful information often comes from real-life interactions with fellow developers, whether it's a brief exchange by the coffee machine, regular tech community meetings within the company, or in-depth discussions about ongoing issues within my project team.
Join us!
Senior software developer
Join us to advance your software development skills and work on diverse tech stacks with a focus on continuous improvement.
