This narrative reflects a conversation with Natalie Gilbert, a 30-year-old data scientist at AT&T, whose father, Mazin Gilbert, worked as a researcher at the company’s Bell Labs division. The interview has been edited for brevity and clarity.
As a child, my understanding of AT&T was quite limited.
My knowledge came primarily from my father, who was involved in speech recognition. He collaborated with pioneers like Yann LeCun, who was innovating handwriting detection, and Dennis Ritchie, the creator of the C programming language.
The foundations of his work in speech recognition and synthesis have been instrumental to my role today in generative AI. The technologies I develop are built on the same principles he pioneered: convolutional neural networks that allow computers to analyze inputs like images and sounds. It’s fascinating to witness the evolution of these foundations.
AT&T
Their groundbreaking discoveries have paved the way for advancements in AI, making these technologies increasingly autonomous.
As a kid, I spent a lot of my afternoons in my dad’s office, witnessing passionate discussions and intricate diagrams drawn on whiteboards.
This exposure inspired me to create my own decision trees, albeit nonsensical ones, which honed my creative and analytical thinking.
One memorable project we collaborated on was called Dr Bot. It was an early version of a large language model designed to evaluate symptoms and suggest appropriate care options.
From Whiteboarding to Coding and Back
My work with AI agents fundamentally revolves around decision trees that help guide pathways from point A to B—an essential lesson I learned from my father early on.
The interaction with humans is vital in developing AI technologies.
In AT&T’s Chief Data Office, we are innovating the approach to HR technology, aiming to simplify how employees resolve HR issues. Our goal is to have an AI agent identify the relevant policies and procedures for each individual situation. This is no small feat in an organization as extensive and intricate as AT&T.
AT&T
In my own work, I utilize a coding assistant that significantly boosts my productivity. However, those creating AI tools still need a solid grasp of the underlying technologies that support large language models and machine learning.
New AI Tools are Amazingly Powerful, but They Can’t Do Everything
As these copilots gain popularity, it’s crucial to understand the fundamental workings of these technologies to avoid pitfalls.
If you aren’t equipped with knowledge about how the code manages edge cases, your AI tools may fall short.
Moreover, it feels like individuals are expected to continuously learn new skills every couple of months.
Today, large language models feel far more conversational and intuitive, requiring me to spend considerable time on prompt engineering. Instead of coding, it often revolves around using natural language to communicate more effectively with machines.
This is somewhat ironic, as it mirrors the work my father was engaged in three decades ago.
Witnessing the transformation of AI throughout my lifetime and carrying forward my father’s legacy feels both surreal and rewarding. I am honored to continue the vital work he started.
