Large Language Models 101 with Michelangelo D'Agostino

Innovation and the Digital Enterprise

Today we’re sharing another insightful presentation from our most recent Innovative Executives League Summit, where Michelangelo D’Agostino, VP of Machine Learning at Tegus, delivered a foundational lesson about large language models. Imagine you are Rip Van Winkle, as Michelangelo puts it, and you have woken up after a long sleep and encountered the current AI landscape. What have you missed? What do you need to know to move forward? Calling upon his data analysis and machine learning expertise, Michelangelo offers clear, concise insights to introduce audiences to the capabilities and shortcomings of large language models today.

In this presentation, Michelangelo integrates large language models to demonstrate their abilities. Defining the term and other critical ones (What does GPT mean?), he dives into the factors that have led to the exponential growth in these models since 2020 and details the training methodologies that led to major advances. Michelangelo covers how instruction tuning brought an exercise in probability to usefulness that will change industries.

Offering insight into the challenges large language models are encountering, Michelangelo walks audiences through a “hallucination,” where the LLM offers a confident answer that is incorrect—a concerning flaw--and displays how prompt engineering generates the correct result with a minor tweak. With the input and output being natural language, Michelangelo encourages people to embrace the low barrier of entry to try out the models directly (OpenAI’s ChatGPT, Anthropic’s Claude, and Google’s Bard) by writing prompts and learning its capabilities firsthand. Michelangelo shares the areas where he’s excited about the potential of large language models and their transformative power for text-heavy industries.

• (00:58) – Demystifying AI
• (03:37) – Large language models
• (05:13) – Unpacking training
• (08:43) – Why now?
• (12:05) – Increased potential
• (14:50) – Hallucinations
• (16:26) – Prompt engineering
• (18:25) – Applications of language models
• (22:48) – Play around with it!

Michelangelo D’Agostino is the Vice President of Machine Learning at Tegus. Previously, he held leadership roles in data and machine learning at Cameo and ShopRunner. Michelangelo’s career as a technologist career is marked by his exploration of large language models and their applications in financial text data. He studied physics, earning a bachelor’s degree from Harvard University and a Ph.D. from the University of California, Berkeley.

If you'd like to receive new episodes as they're published, please subscribe to Innovation and the Digital Enterprise in Apple Podcasts, Google Podcasts, Spotify, or wherever you get your podcasts. If you enjoyed this episode, please consider leaving a review in Apple Podcasts. It really helps others find the show.

Podcast episode production by Dante32.